Planet Caml [EN]

0005622: typing error : The implementation does not match the interface

Thu, 24 May 2012 19:16:50 +0000

I attached the file(ocaml 3.12.1) I have a deadline next week, I would appreicate if anyone could give me some hints. (maybe I am wrong, the type system of Object layer is nontrivial to me)

0005365: Spurious "PIE" warning in Mac OS X 10.7 Lion

Thu, 24 May 2012 17:37:40 +0000

In Mac OS X 10.7 Lion, there's a spurious warning when ocamlscript is compiled, and whenever ocamlscript itself runs. This weird "PIE" warning is specific to Lion, but it appears across languages, in Haskell as well. $ cat hello.ml #!/usr/bin/env ocamlscript -o hello print_endline "Hello World!" $ ./hello.ml ld: warning: PIE disabled. Absolute addressing (perhaps -mdynamic-no-pic) not allowed in code signed PIE, but used in _caml_program from /var/folders/ch/_lv6l5q936n1xv03j5qp28y40000gn/T/camlstartup2188ff.o. To fix this warning, don't compile with -mdynamic-no-pic or link with -Wl,-no_pie Hello World!

0005578: Windows: Exception raised when reading from a socket

Thu, 24 May 2012 17:17:45 +0000

In recent versions of the trunk, the following piece of code raises an exception on input_line: let () = let sock = Unix.socket Unix.PF_INET Unix.SOCK_STREAM 0 in Unix.setsockopt sock Unix.SO_REUSEADDR true; Unix.bind sock (Unix.ADDR_INET (Unix.inet_addr_any, 8090)); Unix.listen sock 10; let (client, _) = Unix.accept sock in Unix.set_close_on_exec client; let ic = Unix.in_channel_of_descr client in print_endline (input_line ic) (compiled with: ocamlc -custom -o test_http.exe unix.cma test_http.ml to trigger the error, just connect to 127.0.0.1:8090 (with telnet, a browser, ...) The exception is: Fatal error: exception Sys_error("Invalid argument") I've tested this with the trunk (msvc and mingw ports). I'm pretty sure this used to work with previous versions. Replacing input_line with Unix.recv seems to work.

0005325: Blocked Unix.recv in one thread blocks Unix.send in another thread under Windows

Thu, 24 May 2012 17:11:42 +0000

It appears not be possible to write to a socket in one thread if another thread is blocked in a [recv] call on the same socket (and presumably vice versa). This behaviour only seems to affect Windows - tried on both Windows 7 with 3.12.0 and Windows XP with 3.10.1 with the same effect.

Fix for PR#5578 left on trunk but removed from 4.00 because of PR#5578.

Thu, 24 May 2012 16:41:34 +0000

m Changes

Fix for PR#5578 left on trunk but removed from 4.00 because of PR#5578.


git-svn-id: http://caml.inria.fr/svn/ocaml/trunk@12481 f963ae5c-01c2-4b8c-9fe0-0dff7051ff02

0005321: Graphics: auto_synchronize does not always work as intended

Thu, 24 May 2012 16:29:17 +0000

The following code shows the bug: open Graphics open Printf let graph s = open_graph(" 200x200" ^ s); auto_synchronize false; moveto 10 10; lineto 100 100; ignore(Graphics.wait_next_event [Graphics.Key_pressed]); synchronize(); ignore(Graphics.wait_next_event [Graphics.Key_pressed]); close_graph() let () = graph ""; (* bug *) graph "-0" (* fine *)

0005616: Please claritfy license and copyright holders of ocamlbuild/manual/*

Thu, 24 May 2012 16:27:58 +0000

Hi, It seems that files shipped under "ocamlbuild/manual/" do not have a declared copyright owner or a license. Admittedly, manual.tex's has a list of authors but they are not the copyright holders (it is not said explicitly). So we cannot redistribute them or do anything useful with them. The LICENSE file says the following: > and "the Compiler" refers to all files marked "Copyright INRIA" in the > following directories and their sub-directories: > > asmcomp, boot, build, bytecomp, debugger, driver, lex, man, > ocamlbuild, ocamldoc, parsing, testsuite, tools, toplevel, typing, > utils, yacc but affected files are not maked "Copyright INRIA" and thus not covered by the LICENSE file. Can you please clarify their license and copyright holders? Besides, it would nice to also provide the ocamlbuild's manual (generated .pdf and .html) in <a href="http://caml.inria.fr/pub/distrib/ocaml-$ver/">http://caml.inria.fr/pub/distrib/ocaml-$ver/</a> [<a href="http://caml.inria.fr/pub/distrib/ocaml-$ver/" target="_blank">^</a>] since it looks like a canonical location for ocaml documentation. Regards,

PR#5616: ocamlbuild's manual is now part of the "official" OCaml reference manual, so just remove these files, it's the easier answer to the copyright-minded folks.

Thu, 24 May 2012 16:26:07 +0000

- ocamlbuild/manual/.ignore
- ocamlbuild/manual/Makefile
- ocamlbuild/manual/manual.hva
- ocamlbuild/manual/manual.tex
- ocamlbuild/manual/myocamlbuild.ml
- ocamlbuild/manual/trace.out
- ocamlbuild/ocamlbuild-presentation.rslide

PR#5616: ocamlbuild's manual is now part of the "official" OCaml reference manual, so just remove these files, it's the easier answer to the copyright-minded folks.



git-svn-id: http://caml.inria.fr/svn/ocaml/trunk@12479 f963ae5c-01c2-4b8c-9fe0-0dff7051ff02

0004653: Install utils/config.cmi so OCaml programs can query their environment

Thu, 24 May 2012 16:19:50 +0000

toplevellib.cma contains utils/config.cmo which contains useful information for OCaml based installers. If utils/config.cmi is installed as well then it would be possible to access this information.

0001804: "make install" should install interface files for toploop-manipulation modules

Thu, 24 May 2012 16:19:50 +0000

Full_Name: Aleksey Nogin Version: 3.06, 3.07beta2 OS: Red Hat Linux Submission from: mojave1.cs.caltech.edu (131.215.44.186) In order to create a custom Ocaml toploop and to be able to manipulate it successfuly, one needs access to a bunch of interface files that are not normally copied into LIBDIR by ocaml's "make install". For example, to be able to compile code that uses Toplevel.execute_phrase, one needs access to parsetree.cmi... Currently (see <a href="http://cvs.metaprl.org:12000/cvsweb/metaprl/editor/ml/shell_p4.ml?rev=HEAD&content-type=text/x-cvsweb-markup">http://cvs.metaprl.org:12000/cvsweb/metaprl/editor/ml/shell_p4.ml?rev=HEAD&content-type=text/x-cvsweb-markup</a> [<a href="http://cvs.metaprl.org:12000/cvsweb/metaprl/editor/ml/shell_p4.ml?rev=HEAD&content-type=text/x-cvsweb-markup" target="_blank">^</a>] for the code of the file that manipulates the toploop) we end up having to manually copy the files: parsing/location.cmi parsing/location.mli parsing/longident.cmi parsing/longident.mli parsing/parsetree.cmi parsing/parsetree.mli typing/typecore.cmi typing/typecore.mli to LIBDIR. Would you please consider modifying the "install" target in the Makefile to copy these files (and may be other interface files needed to take full advantage of Toploop module - such as env.{mli,cmi}, for example) somewhere under LIBDIR? Thanks a lot!

fix two bugs in commit 12453

Thu, 24 May 2012 15:12:37 +0000

m stdlib/hashtbl.ml

fix two bugs in commit 12453

git-svn-id: http://caml.inria.fr/svn/ocaml/trunk@12476 f963ae5c-01c2-4b8c-9fe0-0dff7051ff02

0005621: Crash

Thu, 24 May 2012 12:41:19 +0000

Toplevel crashes as soon as one ask for a few million bytes of memory. See program below.

2012/05/23/Twitter

Wed, 23 May 2012 23:59:59 +0000

02:32:14  <raganwald> “A Haskell newbie is someone who hasn’t yet implemented a compiler. They’ve only written a monad tutorial.” via http://t.co/22Jg0t8M   | 205123944854986752
02:36:16  <planet_ocaml> 2012/05/22/IRC: 00:02:35 <eikke> removing the function call overhead by copying the required Buffer functions i... http://t.co/Sv3fI3KG   | 205124962867085312
09:43:57  <planet_ocaml> PR#5620: invalid printing of type manifest (camlp4 revised syntax): m camlp4/Camlp4/Printers/OCamlr.ml PR... http://t.co/V1R9BqQc   | 205232592902041600
09:43:58  <planet_ocaml> update related to commit 12473 (fixing PR#5620): m Changes update related to commit 12473 (fixing PR#5620... http://t.co/2G2bctso   | 205232593870921728
12:28:29  <planet_ocaml> OCaml for the Masses: Here's one I meant to blog earlier: OCaml for the Masses by Yaron Minsky, from the Nov 201... http://t.co/6u5oiZqS   | 205273997187887104
14:56:32  <stewartadcock> Writing Ocaml code today. Sure it's elegant language, not so sure it's good for my productivity. Makes change from procedural algos though.   | 205311254133342209
15:46:37  <akoprowski> "OCaml for the Masses" by Yaron Minsky, http://t.co/z2lg4mIR   | 205323857509355520
17:04:23  <wagerlabs> OCaml and Haskell are struggling mightily in my head.OCaml, the incumbent,thinks he cannot be unseated.Erlang won over Haskell years ago.   | 205343429222342656
19:09:13  <mentby> Tuareg-mode and pipes | #caml #ocaml http://t.co/XzkNcXQw   | 205374844727083010
19:30:16  <Jose_A_Alonso> OCaml for the masses. (Why the next language you learn should be functional). By Yaron Minsky http://t.co/aeCM9aCU #FP #OCaml   | 205380144439898112
21:36:01  <wtiv770> Practical OCaml (Hardcover): Objective Caml (OCaml) is an open sourced programming language that allows a progra... http://t.co/BQAnZsAE   | 205411786936168448

2012/05/23/IRC

Wed, 23 May 2012 23:59:59 +0000

05:44:15  <IamUnix> Why should i use Caml language not Pascal/C or Go language? What is the main difference?
05:54:48  <IamUnix> Is Caml dead? no innovations anymore.
06:04:45  <adrien> http://caml.inria.fr/cgi-bin/viewvc.cgi/ocaml/trunk/Changes?view=markup
06:42:43  <ssbr> IamUnix: The main difference is probably that caml has a great type system.
10:10:40  <IamUnix> With Caml, is it possible to do C like pointers and pre-processing ?
10:11:26  <adrien> pre-processing, not in the language (like in C's case actually): there's camlp4 and a number of existing syntax extensions for it
10:11:33  <adrien> define "C-like pointers"
10:12:51  <IamUnix> adrien, OK - yes, I mean such: float *px; px = &x;
10:13:01  <adrien> no
10:13:22  <adrien> but if you tell us what you are trying to achieve (and not only "how"), we might help better
10:14:00  <IamUnix> adrien, i am learning Caml, from my C, Java experience, just comparing what i can and what i cant. To quickly catch the basics.
10:16:31  <IamUnix> Can i do something like this ? struct computer {     float cost;     int year;     int cpu_speed;  char cpu_type[16]; }; typedef struct computer SC;
10:17:21  <vincenzoml> Hi there, I'm trying to learn ocsigen. I can't start the web server for the first example in the tutorial:   Fatal error: cannot load shared library dlllwt-unix - Reason: dlllwt-unix.so: cannot open shared object file: No such file or directory
10:18:08  <vincenzoml> the dlllwt-unix.so library is in /usr/local/lib/ocaml/stublibs/dlllwt-unix.so
10:18:11  <djcoin> vincenzoml: there is a specific #ocsigen channel by the way
10:18:17  <vincenzoml> ah thanks, I did not know
10:18:31  <vincenzoml> I'll ask again there
10:18:34  <djcoin> :)
10:19:20  <adrien> it'll be interesting to say which OS you're running and how you're installing everything
10:20:42  <vincenzoml> thanks adrien :)
10:23:39  <IamUnix> Do i have possibility to do such `union` in Caml e.g: union automobile { int year; } sedan; union automobile *pointer; pointer->year = 1997;
10:24:23  <IamUnix> or something: struct survey { int year; union { int year; } provider; }
10:24:37  <flux> iamunix, hmm, your example doesn't really make sense even in C
10:24:50  <flux> you need to have at least two elements you make a union out of for it to make sense..
10:24:51  <adrien> unions, no, because they're unsafe
10:25:01  <flux> but probably you want to use ocaml sum types
10:25:05  <adrien> flux: never read the code of the linux kernel ;-)
10:25:19  <IamUnix> OK
10:25:20  <adrien> but as flux said, there is a different mechanism
10:25:25  <flux> adrien, they have single-field unions?
10:25:47  <adrien> flux: of course :-)
10:26:03  <adrien> socket code
10:26:15  <IamUnix> Can i do such: enum { success, fail, max_len = 81 } , the term `enum` or typedef int twobyte; twobyte i, j; ?
10:27:24  <flux> iamunix, you have type aliases in ocaml. if you want to have constants, you just put a name to them; all variables in ocaml are const by default
10:27:39  <IamUnix> OK
10:28:01  <mfp> IamUnix: the idiomatic equivalent of both unions and enumerations in OCaml is sum types (= disjoint unions)
10:28:02  <flux> probably you want to use sum types for those as well, though, and have a separate way to map them into numbers if you want
10:28:08  <IamUnix> I am trying to language casting. To understand quickly. nvm.
10:28:26  <flux> "success, fail, max_len" is an example of code ocaml probably wouldn't have
10:29:26  <IamUnix> Can i do goto stuff such as in C: label-name:   statement1;   statement2;  goto label-name;
10:29:43  <flux> there is no goto in ocaml
10:30:00  <flux> I think it's time for you to introduce to some ocaml tutorials and documentation ;)
10:30:13  <flux> (s/introduce/introduce yourself/)
10:32:53  <IamUnix> OK - Thank you.
10:33:33  <djcoin> Goto statements
10:33:35  <djcoin> lol
10:40:01  <IamUnix> NVM, just casting language for better/quick catch. Can i do Java like mapping to any object? e.g: private static Map<String, JTextField> currentComponents = new HashMap<String, JTextField>();
10:41:12  <djcoin> IamUnix: do you really think java is close to OCaml ?
10:41:12  <IamUnix> OR try { ... } catch (IOException io) { io.dump... ; }
10:41:26  <IamUnix> NO just the terms getting used.
10:43:14  <IamUnix> Those type of terms: private static Map<String, JTextField> currentComponents = new HashMap<String, JTextField>(); OR private static List<JTextField> list = new ArrayList<JTextField>(); OR type cast JTextField goal = (JTextField) (JTextComponent) (Object) "serverTextField";
10:43:39  <IamUnix> Map or List or multiple cast from right to left...
10:43:49  * ski is not sure what IamUnix is asking about
10:44:17  <IamUnix> OCaml cheat-sheet for background of C/Java users.
10:44:39  <ski> i'm not sure sure a cheat-sheet would make much sense ..
10:44:46  <djcoin> IamUnix: what don't you just send the whole program you are trying to convert line by line ? :)
10:44:54  <ski> .. except to tell you to forget most of what you know from C or Java :)
10:45:17  <ski> IamUnix : there are no casts in O'Caml
10:45:26  <IamUnix> OK
10:46:36  <ski> (you can convert from a subtype to a supertype (but not vice versa), though)
10:47:30  <ski> IamUnix : if you're looking at a general introduction, maybe you'd like to take a peek at <http://caml.inria.fr/pub/docs/manual-ocaml/manual003.html> ?
11:07:48  <mrvn_> -c/etc/wpa_supplicant.conf
11:08:38  <mrvn_> ups, ewin
13:24:14  <everyonemines> IamUnix: your goto example would be: let rec loop() = statement1;  statement2; loop()
13:24:21  <everyonemines> or for a downward goto, an exception
13:24:38  <everyonemines> ocaml exceptions are fast enough that you can use them like that
13:25:20  <djcoin> goto are such an essential programming feature and best practice :>
13:26:02  <IamUnix> Thanks!
13:26:30  <everyonemines> if you want to map, you do it like
13:26:36  <everyonemines> Array.map [array]
13:26:44  <everyonemines> er, Array.map [function] [array]
13:26:52  <everyonemines> read the module reference for the standard library functions
13:27:03  <IamUnix> OK
13:27:31  <everyonemines> if you want to do lots of array stuff, you can "open array" then map function array
13:27:55  <everyonemines> (but remember that modules must be capitalized)
13:28:02  <Ptival> "open Array"
13:28:44  <everyonemines> or, Array.(map function array) opens it locally
13:29:16  <everyonemines> once you get to know ocaml I think you'll like it better than java :-)
13:29:30  <IamUnix> I hope so :)
13:29:42  <everyonemines> what kind of thing are you working on?
13:29:56  <everyonemines> or is it class stuff?
13:30:20  <IamUnix> Text translations
13:30:23  <Ptival> (Java is always "class" stuff :p)
13:30:44  <IamUnix> Chinese to Bengali or reverse etc...
13:30:49  <djcoin> class stuff, rofl :)
13:30:52  <everyonemines> That's a pretty tough problem.
13:30:57  <f[x]> classy!
13:31:03  <everyonemines> but I like algorithmic stuff :-)
13:33:35  <everyonemines> IamUnix: Have you done natural language processing stuff before?
13:35:20  <IamUnix> everyonemines, Not yet. Textual translations right now for one project
13:37:30  <everyonemines> well, you can learn a lot from working on that, and ocaml is useful, but if you need something now, you might be better off using the google translate api :-/
14:05:30  <mfp> according the the gc.mli,   let v = ... in Gc.finalise (fun x -> ...) v    "will not work as expected" because "anything reachable from the closure of finalisation functions is considered reachable"
14:06:46  <mfp> does that only apply to the case where ... captures v, or is it general? (and if so, why does the closure enclose the whole environment?)
14:23:47  <thelema> mfp: it only makes sense to me that it's when ... captures v
14:26:44  <mfp> hmm maybe there could be "surprise captures" in cases like   (fun () -> .... a .... b .... Gc.finalise (fun _ -> ... a .... b ...) c)
14:27:40  <mfp> in some hypothetical OCaml implementation (AFAIK not the current one) that reused the environment for the outer closure in the inner one
14:32:36  <thelema> probably just an over-warning; if people always define their finalization function before declaring their value...
14:32:52  <mfp> guess so
15:28:55  <hcarty> orbitz: The upstream PCRE file is missing.  I sent the author an email asking about it.
15:39:27  <ssbr_> Which of these is generally considered more idiomatic? "let x y z = match z with ..." or "let x y = function ..." ?
15:41:24  <hcarty> ssbr_: I'm not sure which is generally considered more idiomatic, but I use the first when there are multiple arguments and the second when there is only one.
15:41:58  <ssbr_> Yeah. I just don't know if this currying stuff is _actually_ confusing, or if it's just something I'm not used to yet, you know?
15:42:18  <ssbr_> It's difficult to edit this codebase because it was written by two people: one who was really good at ocaml, and one who was really bad.
15:42:26  <ssbr_> (worse than me, and I just started, really.)
15:43:01  <ssbr_> This is not confusing though, so it's not like that this time. I'll leave it as is.
15:43:37  <hcarty> ssbr_: That's why I have that split - I find it a bit easier to read the resulting code as a human.
15:44:31  <hcarty> When there are multiple arguments I have found that throwing "= function" on the end makes it more difficult to parse the function at a glance.
15:45:47  <ssbr_> Yeah, that's what I figured.
15:46:19  <ssbr_> Hm. Reading the body of this function, it might have been written by the bad programmer. But the good programmer is the one that loves currying everything...
15:46:26  <ssbr_> Aieie
15:47:04  <ssbr_> By the way, what is "||" in ocaml?
15:47:28  <ssbr_> errr, wait, I think that's [||]
15:47:32  <ssbr_> is that an empty array?
15:48:25  <Qrntz> ssbr_, yes, [| |] is an empty array
15:48:35  <Qrntz> the amount of whitespace between the braces is arbitrary
15:49:18  <Qrntz> «||» is a logical OR
15:49:35  <ssbr_> Right, I realized my question was ambiguous slightly too late. :(
15:49:51  <ssbr_> (and that the braces were probably part of the syntax -- the return type was array)
15:50:04  <Qrntz> [ || ] wouldn't make sense
15:50:07  <Qrntz> so, perhaps
15:50:21  <ssbr_> Thanks :)
15:50:40  <Qrntz> you're welcome
15:50:59  <f[x]> [(||)]
20:02:37  <osa1> can anyone recommend me an introduction text for ocaml? I have functional programming(haskell, scheme) experience but I'm not an expert in types and advanced functional programming techniques
20:03:53  <bitbckt> osa1: www.cs.caltech.edu/courses/cs134/cs134b/book.pdf
20:04:12  <bitbckt> that's probably the most approachable intro.
20:04:26  <osa1> bitbckt: thanks
20:04:33  <bitbckt> np
20:04:48  <bitbckt> the official manual is thorough, if a bit dry.
20:04:52  <bitbckt> best used as a reference.
20:05:57  <osa1> bitbckt: oh, it's from author of real world ocaml ..
20:06:08  <bitbckt> one of them, yes.
20:35:45  <orbitz> hcarty: thanks
21:01:51  <wmeyer``> ssbr: Abuse currying that's what it was designed for. Beware combinators like "flip" and point free coding however. In terms of let foo a b c = ... then 1) Usually my functions are single argument and the rest is the environment outside a closure - but I'm doing a lot of AST transformations - if they have second argument - then it's accumlator or environment - in case of API situations might be very different! 2) If they acc
21:01:51  <wmeyer``> than one and you need to pattern match for the last argument - I usually use "function", that's for me more idiomatic. In general what type checks and is readable in 80% is enough.
21:09:53  <wmeyer``> thelema_: I rebased my branch - now you can pull
21:15:04  <ski> (possible cut off "... 2) If they acc")
21:54:55  <osa1> is there a way to load a source into ocaml repl? like :load in haskell's ghci
21:55:11  <gildor_> #load ?
21:55:28  <Qrntz> #use "file.ml" ;;
21:56:07  <osa1> thanks
21:56:40  <Se7en> yes, use that
22:47:27  <orbitz> Does the run time gurantee the order modules are executed in, fo rexample can I have a module register itself with a plugin registry like let () = Registry.register yadda;; or could it be possible Registry has not ben initialized yet?
22:48:23  <_habnabit> orbitz, if you're referencing Registry, then it's been initialized
22:48:45  <_habnabit> orbitz, initialization order is based on which modules depend on each other
22:50:37  <orbitz> Excellent
22:50:48  <orbitz> Is what I am trying to do a good idea?
22:51:07  <_habnabit> it doesn't seem like a bad idea. the ocaml stdlib already does this.
22:51:12  <orbitz> Basically I want a factory I guess
22:51:25  <_habnabit> this doesn't seem like a factory
22:51:30  <Qrntz> I would be very interested in something like that
22:51:32  <orbitz> Being able to handle multiple database types at runtime, for example
22:51:35  <Qrntz> (a plugin factory)
22:51:41  <_habnabit> it's a registry
22:52:01  <orbitz> _habnabit: Well, I'd register with it, then when I want a new db connection i'd say Registery.get db_type
22:52:14  <orbitz> and then I'd have a record (or something?) that has functiosn to play with a db
22:52:34  <_habnabit> if you're using 3.12, you could use modules-as-values
22:53:11  <orbitz> Yeah, I don't think that necessarily changes much though
22:53:16  <orbitz> I'd still eb doing a .get somewhere, right?
22:53:31  <_habnabit> well, yeah
22:53:38  <_habnabit> I just mean instead of a record
22:53:45  <orbitz> Yeah
22:54:04  <orbitz> I'd probably have get really be .connect or soemthign and give me back something I can use a Connection module on
23:09:03  <wmeyer``> orbitz: with first class values you can package them and assign to mutable variable during load
23:09:22  <wmeyer``> orbitz: s/values/modules/

0005620: camlp4 printer bug

Wed, 23 May 2012 17:46:53 +0000

a minimal example: camlp4rf -str 'type t = M.t == [A|B];' -printer r type t = M.t = [ A | B ];

OCaml for the Masses

noreply@blogger.com (Philip Wadler) — Wed, 23 May 2012 10:26:00 +0000

Here's one I meant to blog earlier: OCaml for the Masses by Yaron Minsky, from the Nov 2011 CACM. Is material on FP for developers the new trend?

update related to commit 12473 (fixing PR#5620)

Wed, 23 May 2012 08:23:52 +0000

m Changes

update related to commit 12473 (fixing PR#5620)

git-svn-id: http://caml.inria.fr/svn/ocaml/trunk@12474 f963ae5c-01c2-4b8c-9fe0-0dff7051ff02

PR#5620: invalid printing of type manifest (camlp4 revised syntax)

Wed, 23 May 2012 08:19:46 +0000

m camlp4/Camlp4/Printers/OCamlr.ml

PR#5620: invalid printing of type manifest (camlp4 revised syntax)

git-svn-id: http://caml.inria.fr/svn/ocaml/trunk@12473 f963ae5c-01c2-4b8c-9fe0-0dff7051ff02

Students at Gallium

Wed, 23 May 2012 08:00:00 +0000

A short presentation of the various young people that regularly gather (or used to gather) at the conversation-rich and often surprising coffee breaks we have at Gallium.

Onsite Gallium PhD students

In decreasing order of elderness:

Jonathan Protzenko had the idea of this blog. He is working with François Pottier on programming language approaches to have finer control of mutability and aliasing. He also hacks for Mozilla on his spare time (eg. he's the author of the world-famous Conversations extension for Thunderbird) and even wrote a full book about XUL in his young age, before seeing the truth of functional programming. From these fond experiences he's formed a strong sense of contribution to the community, which makes him an asset of the OCaml bug triaging volunteer squad. Let's just add that he took responsability for the Windows installer, to give you an idea of his sense of self-sacrifice. His is also in charge of the "junior seminar" here at INRIA Rocquencourt, a great opportunity for students to practice their communication skills, that I hope this blog will complement.
Fixing Windows bugs is rarely a charming activity, and Jonathan's mood sometime pays for it. Julien Cretin is the ever-smiling one. He works on coercions (that is, in his mind, mostly anything typing-related) with Didier Rémy but is in fact busy preparing the revolution: anytime soon programmers will stop using term syntax and variables, and instead draw graphs. He's also an asset for our cultural diversity, as his main working language is Haskell; he had the luck to do an internship in MSR Cambridge last year, where he helped a dream team to promote Haskell (PDF) through kind-level programming.
I'm Gabriel Scherer and I should be working on module systems, also with Didier Rémy. In fact, I'm trying to recover from a serious illness: the inability not to answer a question on the Internet -- or comment on a good post. Please also don't approach me with interesting research topics (anything programming language related), I'm much too likely to get distracted.
Jacques-Henri Jourdan started as a PhD student just a few weeks ago. He has already worked in Gallium as an intern, on a formal correctness proof from a grammar description (in Coq, automatically translated from a Menhir grammar) to a LR automaton. He will now work on mechanized formalizations of abstract interpretation techniques.

Stranger beings

Alexandre Pilkiewicz found that writing a PhD thesis was too easy, and decided to simultaneously start a job at Google. Here, he was working on formalizing compiler optimizations based on polyhedral loop analyses. He is very comfortable with Coq, to the point that he once did a formal proof of correctness for a small patch I proposed for the standard Map data structure, just for fun.
Valentin Robert worked as an intern with Xavier Leroy on Compcert, first on a certified pointer analysis pass, and then on test programs to verify the work of the external non-verified toolchains that take Compcert proved-correct assembly output and generate a real executable. He is our meme expert, and notoriously indecisive. After much hesitation, he will be leaving us in a few months to do a PhD in San Diego.
Raphael Proust is the most recent intern in the team; he will be working with Xavier Leroy on the intermediate representations (in a compiler) of functional languages : CPS, SSA, ANF etc. He previously worked with the Mirage team in Cambridge, doing OS-level development in OCaml, and will probably go back there to start a PhD in a few months. That's a very interesting project with not-so-surprising ties to programming language research; for example he worked on the use of Limel (PDF), a programming language with linear type, to faithfully encode the constraints on low-level dataflow tasks in a networked system.

Recent former students

I won't list all the former students of Gallium, but here are those I had the pleasure to meet while they were still on duty.

Nicolas Pouillard is our former "cultural diversity" leader, using Agda as his proving tool of choice. Coq and OCaml are more in the culture of the place, but anyone is free to choose -- and XMonad is certainly king among the window managers used. Nicolas worked on the hairy subject of representing and formally reasoning on syntactic terms with binders. He was also our local expert on practical uses of secure design or cryptography, such as security by capabilities (a hobby I share), web privacy, BitCoin, Tahoe-LAFS... He is now a post-doc at ITU Copenhaguen as part of the demtech project on "trustworthy democratic technology".
Tahina Ramananandro worked on the craziest topic one could find. What is the hardest way to describe a programming language? I would say it is writing mechanized formal semantics for it. What is the most insanely complex language a twisted mind could want to describe? C++. At the end of a truly heroic effort, Tahina had a formalized semantics of memory layout of structs/classes and construction/destruction of objects. He's now doing a post-doc in the FLINT group at Yale.
Benoît Montagu worked on module systems with Didier Rémy. He studied several ways to enhance Fω with some orthogonal features each adding some aspect of the ML module system. He's now a post-doc in Philadelphia, as part of the quite exciting CRASH/SAFE project. Take a bunch of type system and formal proof experts, and ask them to design an end-to-end (hardware and software) system aimed for security, safety and reliability; they produce a dynamic language.

2012/05/22/Twitter

Tue, 22 May 2012 23:59:59 +0000

01:40:38  <yang_yihming> vw009: Which language do you often use in parallel programing? Ocaml? Erlang? C01? Or some other language?   | 204748571533258754
07:21:45  <planet_ocaml> OCaml-MySQL 1.1.1: A package for Objective Caml that provides access to MySQL databases. http://t.co/jMyNorhR   | 204834418655891456
07:21:45  <planet_ocaml> ocaml-snappy 0.0.1: Bindings to snappy - fast compression/decompression library. http://t.co/AI5IB3fG   | 204834419641548800
07:21:46  <planet_ocaml> ocaml-lzo 0.0.1: Bindings to LZO - a portable lossless data compression library. http://t.co/5F2Sz60z   | 204834420677554176
07:21:46  <planet_ocaml> ocaml-winsvc 0.0.1: The library to make OCaml program act as a Windows service. http://t.co/hZtT9hfZ   | 204834421625454592
07:21:46  <planet_ocaml> RegStab 2.0: RegSTAB is a SAT-solver able to deal with formula schemas: you can give it a scheme o... http://t.co/gO5W6Klx   | 204834422573371392
09:36:07  <planet_ocaml> Iterating over the AST: Context A facetious colleague who claims that he should be better writing his thesis bu... http://t.co/O8yJABB1   | 204868230702051328
10:02:35  <reddit_haskell> A proposal for Open Extensible Datatypes in OCaml, inspired from Löh&Hinze http://t.co/UZNg0HZ5   | 204874890795417600
11:31:32  <debasishg> Patch for extensible open data types in OCaml as a solution to the expression problem .. https://t.co/psWzQAX6   | 204897278404734976
12:20:42  <planet_ocaml> Caml Weekly News, 22 May 2012: Any library for reading/writing compressed files? / extunix 0.0.4, ocaml-mysql 1.... http://t.co/7Y82VvFq   | 204909651609915392
13:31:47  <educamedia> Codepad:online compiler/interpreter for C, C++, D, Haskell, Lua, OCaml, PHP, Perl, Plain Text, Python, Ruby Scheme &Tel http://t.co/q5UCIcBm   | 204927541818032128
16:47:39  <4r_creativita> I no longer count the number of ladies who proposed right after discovering that I was an OCaml programmer.   | 204976829344395265
21:50:19  <KanzariyaJayesh> #teamletsnurture Technical Details: Static Blogging in OCaml with Stog http://t.co/0hfvQAat   | 205053001495347201

2012/05/22/IRC

Tue, 22 May 2012 23:59:59 +0000

00:02:35  <eikke> removing the function call overhead by copying the required Buffer functions in a module in the same file, then compiling with agressive inlining, helps +- 10% (asm shows all call's are gone)
03:00:21  <ssbr> I figured out what I want!    I definitely wish I could do "match x with ((Foo | Bar) (x, y, z)) -> ..."
03:00:57  <ssbr> hm, but then I'd want a label for whatever Foo | Bar was so that I could return a new Foo/Bar as necessary...
03:01:04  <_habnabit> match x with Foo (x, y, z) | Bar (x, y, z) -> x + y + z
03:01:28  <ssbr> _habnabit: that is annoying. Also I'd forgotten my actual problem
03:01:51  <ssbr> which is Foo (x,y,z) -> blah x y z | Bar (x,y,z) -> blah x y z
03:02:09  <_habnabit> and why couldn't you write it the way I said?
03:04:22  <ssbr> _habnabit: oh, because it's really annoying
03:05:20  <ssbr> _habnabit: it ends up looking like this: match x with Foo (x, y, z) -> let renamed = (rename_params x z) in ((rename_values y renamed), renamed) | Bar (x, y, z) -> let renamed = (rename_params x z) in ((rename_values y renamed), renamed)
03:05:42  <_habnabit> er, I don't think you read what I said
03:05:45  <ssbr> I guess I should write that as a function? makes it spaghetti-like
03:05:56  <ssbr> _habnabit: ehhhhhh! I forgot the last bit
03:05:59  <_habnabit> note how mine only ahs one ->
03:06:27  <ssbr> _habnabit: I think it was actually ((rename_values y renamed), Foo x y renamed) or something like thbat
03:06:35  <ssbr> Foo (x,y,renamed)
03:06:40  <ssbr> I really wish I had the code in front of me
03:06:47  <_habnabit> oh, well
03:06:53  <ssbr> ... crap, I mistyped my "actual problem"
03:06:55  <_habnabit> then I'd write a function, yeah
03:07:03  <ssbr> _habnabit: sure, except... how?
03:07:25  <ssbr> Either way I _need_ that (let renamed = ... in ...) bit to be copy-pasted
03:07:43  <_habnabit> it is kind of a problem that you can't pass around a type constructor like a function
03:07:57  <_habnabit> I'd like to see the actual code, though :p
03:08:08  <ssbr> _habnabit: I'll show it to you tomorrow
03:08:40  <ssbr> I might be misremembering  _this_ particular annoying case, although there were others that were less horrible
03:09:13  <ssbr> anyway, um, yes, I communicated poorly. Hush, I just woke up. :<
03:45:06  <wmeyer``> :)
03:45:31  <wmeyer``> ssbr: lucky, i am going to bed now
03:45:50  <ssbr> wmeyer``: hardly lucky. I missed a whole day
03:45:52  <ssbr> it is midnight here
03:46:35  <wmeyer``> ssbr: that's exactly what I did yesterday
03:46:51  <wmeyer``> ssbr: woke up at midnight and started hacking
03:48:16  <wmeyer``> ssbr: so where are you at, if it's midnight I presume is some way half way to states from UK
03:49:25  <ssbr> wmeyer``: Toronto!
03:49:35  <wmeyer``> ssbr: that's pretty cool :)
03:49:41  <ssbr> it's actually really hot
03:49:43  <ssbr> and I don't have AC
03:50:10  <wmeyer``> ssbr: me either - but here is not needed
03:50:22  <ssbr> Lucky. :)
03:50:59  <wmeyer``> ssbr: oh, I wish the weather was slighly better - but anyway most of the I spent in the bunker so I don't mind
03:51:41  <wmeyer``> ssbr: but the beer festival has just started
03:51:58  <ssbr> Oof. So you're missing that then?
03:51:58  <wmeyer``> ssbr: today \o/ so I might come out to drink some pints
03:53:38  <wmeyer``> sometimes I wish the weather was better yes, not that often, in fact i am mostly fine with wearing almost whole year a light jacket
03:54:32  <wmeyer``> ssbr: Camlp4 let me down again :(
03:54:53  <ssbr> hahaha
03:55:00  <ssbr> camlp4 always lets me down :<
03:55:08  <ssbr> I keep expecting it to be able to grok what it does not grok
03:55:18  <wmeyer``> ssbr: true, true. :D
03:55:32  <wmeyer``> but this time I had another weird problem
03:56:24  <wmeyer``> on saturday night It was generating an additional leading bar in the generated polymorphic variant types
03:56:52  <wmeyer``> which is strangely fine when you generate OCaml text file
03:57:07  <wmeyer``> but is not when it performs direct marshaling to the compiler
03:57:31  <wmeyer``> so it compiles fine - but when you use -pp it gives weir error message
03:57:47  <wmeyer``> so I work arounded that by using custom ocamlbuild plugin
03:58:03  <wmeyer``> today there is some problem I've never had before
03:58:19  <wmeyer``> it generates: let alu stream = fun _ -> raise (Match_failure ("ghost-location", 1, 0))
03:58:19  <wmeyer``> 
03:58:26  <wmeyer``> very helpful isn't it?
03:58:33  <ssbr> mmm.
03:58:53  <wmeyer``> I am sure however I handle all the cases
03:58:59  <wmeyer``> it was working before the refactoring
03:59:17  <wmeyer``> ups - maybe the list is empty!
03:59:37  <wmeyer``> so it will generate and empty <:expr<>> or whatever
03:59:42  <wmeyer``> and camlp4 goes mad
04:00:26  <wmeyer``> anyway I pretty much done for today
04:00:44  <wmeyer``> time to take a nap before i start doing C++ :/
04:02:52  <ssbr> Goodnight
04:02:57  <wmeyer``> night!
13:39:19  <leopard2> Hi anyone around?
13:39:44  <Drakken> yep
13:40:41  <adrien> sorry, /me is a square
13:40:41  <leopard2> Error: Unbound value get_full_path
13:41:02  <adrien> leopard2: you should pastebin the full code ;-)
13:41:24  <Drakken> get_full_path isn't defined in that scope.
13:41:30  <leopard2> Just running a test on Leopard   prerr_endline ( get_full_path "/usr/lib/haxe/std/StdTypes.hx");;
13:42:22  <Drakken> where is gfp defined?
13:43:37  <leopard2> I was just running a test for Nicolas for haxe issue on leopard just going to try on lion.  But newbie to ocaml.
13:44:25  <leopard2> should the line be different?
13:44:53  <leopard2> this is the compile line... ocamlopt -o test -cclib libs/extc/extc_stubs.o -cclib -lz libs/extc/extc.cmxa test.ml
13:47:36  <adrien> get_full_path is probably from extc
13:47:42  <adrien> so, considering the name
13:47:51  <adrien> try: "Extc.get_full_path"
13:49:55  <f[x]> live telepathy seance!
13:50:04  <leopard2> Thanks adrien  is the compile info correct for mac?
13:50:54  <leopard2> its complaining about Unbound Module Extc
13:52:32  <leopard2> could it be that may ocaml is via mac ports and the poth to extc is wrong?
13:52:44  <leopard2> path
13:52:52  <adrien> could be; it's very hard to say since we have little infos
13:53:12  <adrien> what have you installed? ocaml only? ocaml+haxe? something else?
13:55:32  <leopard2> I have haxe installed but currently a broken haxe version on leopard only.  I have tracked down the break to rev 4188 haxe.  And I have ocaml 3.12.1
13:57:10  <leopard2> its line 1300 or there abouts of typeload.ml this line     mfile = Common.get_full_path file;  that breaks leopard.
13:58:00  <leopard2> removing Common.get_full_path and it works fine. So just running a test Nicolas ask me to run.  But it's not compiling on Lion on leopard.
14:00:54  <leopard2> ok I have...
14:01:00  <leopard2> locate extc_stubs.o
14:01:01  <leopard2> /Users/justinmills/ocaml/extc/extc_stubs.o
14:01:03  <leopard2> /haxe29/haxe-read-only/doc/haxe/libs/extc/extc_stubs.o
14:01:04  <leopard2> /haxe29/haxe-read-only/libs/extc/extc_stubs.o
14:01:06  <leopard2> /usr/lib/haxe/temp/haxe-read-only/libs/extc/extc_stubs.o
14:01:07  <leopard2> /usr/lib/ocaml/extc/extc_stubs.o
14:02:18  <leopard2> I presuem I should change the compile to make sure extc_stubs.o is found
14:02:25  <leopard2> presume
14:04:34  <leopard2> I have changed the compile line to... ocamlopt -o test -cclib /usr/lib/ocaml/extc/extc_stubs.o -cclib -lz /usr/lib/ocaml/extc/extc.cmxa test.ml
14:06:23  <adrien> and?
14:07:54  <leopard2> it is still unbound with or without Extc
14:08:07  <leopard2> what is -cclib doing?
14:08:25  <adrien> system linker arguments
14:08:35  <adrien> tells it to use "-lz"
14:09:20  <leopard2> Are there path differences between leopard and snowy in ocaml ?
14:12:13  <leopard2> by the way is there a good book on ocaml, I can never find any at any book stores ( and tried London, Bath, Oxford ).
14:13:36  <adrien> there's the o'reilly book; now free of charge but not printed
14:13:42  <adrien> and see ocaml-lang.org
14:13:54  <adrien> as for your command, what's the error message for the last one?
14:14:12  <leopard2> is the o'reilly book up to date?
14:14:29  <leopard2> ocamlopt -o test -cclib /usr/lib/ocaml/extc/extc_stubs.o -cclib -lz /usr/lib/ocaml/extc/extc.cmxa test.ml
14:14:30  <leopard2> File "test.ml", line 1, characters 16-29:
14:14:32  <leopard2> Error: Unbound value get_full_path
14:14:42  <leopard2> when I exclude Extc
14:14:47  <leopard2> and
14:15:04  <leopard2> File "test.ml", line 1, characters 16-34:
14:15:05  <leopard2> Error: Unbound module Extc
14:16:48  <adrien> well, for get_full_path, ask Nicolas what should be the line; it's a library function but you're missing the full path probably
14:16:52  <adrien> he should know it
14:17:00  <leopard2> I tried cd /usr/lib/ocaml/extc/   and it takes me to the correct place
14:18:02  <leopard2> I will have to push back on the email yes I just thought I should try IRC often a good way to get more insight into a language.  Thanks for the help anyway.
14:19:31  <adrien> the issue is that it uses a library that most of us don't know about
14:19:36  <adrien> ("extc")
14:24:05  <leopard2> how do I check where my Ocaml is running from?
14:24:17  <adrien> ocamlc -where
14:24:26  <adrien> it shows the standard library path
14:24:38  <adrien> but you really should not have two installations of ocaml available at the same time
14:26:02  <leopard2> ocamlc -where
14:26:03  <leopard2> /opt/local/lib/ocaml
14:27:49  <leopard2> there is no extc folder in there though can I macport it
14:28:47  <adrien> ls /usr/bin/*ocamlopt*
14:30:01  <leopard2> well mine is in usr/lib/ocaml/ does that differ between leopard and snowy?
14:37:16  <adrien> well, not really a matter of mac os x version, more of macports
14:39:57  <leopard2> Older versions of haxe compiled fine so using macports ocaml should not be a problem or atleast haxe source should no mind
14:40:09  <leopard2> not mind
15:18:49  <fpz> window balance
15:20:42  <orbitz> hcarty: Hello!
16:01:20  <orbitz> hcarty: No mor String.sub error, but pcre is 404: failed to get http://oasis.ocamlcore.org/dev/odb/stable/pkg/pcre-6.2.3.tar.gz
16:45:50  <tchell> so I have a neato 20 line program that segfaults.
16:45:54  <tchell> http://hastebin.com/cuwulorogi.coffee
16:46:06  <tchell> And it segfaults in a different place every time, as judged by printf debugging.
16:47:35  <tchell> It opens and closes lots of filehandles, uses Pcre, reads a 22GB file line by line... I'm running out of ideas.
16:47:44  <mfp> tchell: looks like a stack overflow
16:47:54  <tchell> because of my mutually recursive functions?
16:47:55  <mfp> next is not tail-recursive
16:48:05  <tchell> I see.
16:48:17  <mfp> you are calling read inside a try ... with Not_found -> ...
16:48:45  <tchell> ok, I can fix that.
16:49:12  <mfp> the typical idiom to handle that is  let ms = try Some (Pcre.extract ...) with Not_found -> None in  match ms with Some x -> ... | None -> ...
16:49:24  <mfp> allowing you to call perform the recursive calls in tail position
16:49:41  <mfp> -call
16:54:04  <tchell> Ok.
16:54:30  <tchell> Great! That seems to work!  Thanks!
16:54:44  <mfp> np
16:56:39  <wmeyer``> thelema: are you there?
16:58:23  <wmeyer``> thelema: I was thinking currently about parsing _oasis dependencies, I will push another fix for the oasis installation first. BTW: Shouldn't odb have interactive mode?
17:00:22  <wmeyer``> thelema: Another idea -- why to clober user with the unneeded log -- spawn the build, check what was the status code, if it's not 0 then be suspicious and print the accumulated log. In general I would expect that odb would be more elegant in terms of the dialog with the user.
17:42:40  <jaxtr> ahh it's a wonderful day
17:46:46  <wmeyer``> thelema: Just commited the final fix for getting oasis if it's needed
18:33:18  <thelema> wmeyer``: I'll take a look at it now, but I'm about to go, so unless it pulls cleanly, I don't think it'll get merged right now
18:37:49  <thelema> I worry that an interactive interface for odb is...  unnecessary.  I think we can put a program with an interactive interface in the odb repo, but I want simple `curl odb.ml | ocaml` to work out of the box
18:44:05  <wmeyer``> thelema: No worry it can wait, thanks.
18:53:16  <wmeyer``> thelema: I think having some odb-simple or odb-interactive that will call odb.ml put *somewhere* (biggest chalenge) would be an interesting experiment - worth to try - if ocamlbrew supported that - this would be even better. I think we could then make odb.ml better at the same time - sort of separating dialog from the rest of code. So far I am happy on incremental improvements
19:59:06  <ssbr_> I never want to see a fold for the rest of my life
19:59:33  <ssbr_> 100% of the List.fold_lefts and List.fold_rights I've read, I've rewritten as List.map
20:01:20  <adrien> # List.fold_left (+) 0 [ 1; 2; 3; 4 ];;
20:02:02  <adrien> also, if you've rewritten a fold_left into something with map, you've lost tail-recursivity
20:02:11  <jonafan> let List.map f xs = List.fold_right (fun x xs -> f x :: xs) xs []
20:03:03  <jonafan> map and fold are for different things, and you probably should have used map in the first place
20:04:49  <adrien> if you manager to make something using fold_left and a lower complexity than it'd take with map or rev_map, then definitely use fold_left
20:08:20  <ssbr_> adrien: they're implementing map in terms of folds.
20:09:20  <ssbr_> List.fold_right (fun k lst -> (foo k) :: lst) somelist []
20:11:07  <ssbr_> I just replaced something in exactly this form.
20:13:13  <jonafan> well, at least you didn't use a list reference with List.iter
20:13:39  <ssbr_> There's also the thing where I really don't like reading folds -- the associativity issues disturb me out of a relaxed reading :S
20:13:54  <_habnabit> jonafan, fwiw it's legacy code not written by ssbr_
20:14:14  <jonafan> ah
20:14:26  <ssbr_> I'm not sure if I'll grow out of disliking folds. But in this case I think the folds are pretty objectively wrong.
20:14:52  <ssbr_> maybe you can quibble about the fold_lefts that are equivalent to a reverse map?
20:14:59  <ssbr_> I don't remember if there were any of those
20:15:57  <_habnabit> ssbr_, what's the alternative to a fold, though? are you suggesting that map not be implemented in terms of a fold, or what
20:17:07  <ssbr_> _habnabit: I
20:17:11  <ssbr_> whoops
20:17:25  <ssbr_> _habnabit: I don't really care how map is implemented. But I find recursion easier to read.
20:17:41  <ssbr_> Apparently functional programmers hate recursion and love HOFs, but...
20:17:52  <ssbr_> I dunno, fold isn't a HOF I like, anyway.
20:17:58  <ssbr_> Maybe I'll grow out of it.
20:18:12  <_habnabit> that's fine for the case of lists, maybe. but what about folding over a tree?
20:18:22  <_habnabit> you don't necessarily want to expose the implementation
20:19:00  <ssbr_> _habnabit: I've never folded over a tree.
20:21:48  <everyonemines> fold doesn't parallelize, map does
20:22:04  <everyonemines> (but in the case of ocaml map is implemented as a for loop)
20:22:24  <_habnabit> (only if you're mapping a pure function)
20:22:33  <everyonemines> (but you usually are)
20:23:59  <ssbr_> I don't care about CPU usage or stack usage one bit with this code.
20:24:18  <ssbr_> Everything being folded that I've replaced is really tiny.
20:28:29  <wmeyer``> ssbr_: I use a lot of fold, I prefer combinators over explicit recursion
20:29:10  <ssbr_> I dunno. I feel like I should.
20:29:24  <ssbr_> I think after enough of it I will just pattern match the associativity instantly
20:29:38  <ssbr_> as it is I have to think, "OK, fold_right replaces cons with f, and fold_left goes left to right"
20:29:52  <ssbr_> and work it out. Annoying. :(
20:30:10  <ssbr_> (fortunately for stuff like (+) and (*) it doesn't matter at all!)
20:30:22  <Ptival> I just think of it as "ok, need to do something on that list, passing around that result being built..."
20:30:44  <ssbr_> Ptival: you're thinking of constructing a fold though
20:30:49  <ssbr_> I'm thinking of reading. :S
20:31:24  <ssbr_> I guess it applies both ways
20:31:47  <Qrntz> on a slightly related note, would it make more sense to use a string map instead of a (string, string) hash table if 1) I have a small dataset 2) the structure is often fold 3) the structure stays constant throughout the program run?
20:32:09  <Ptival> why would you not use a map?
20:32:16  <Qrntz> I am mostly concerned with performance
20:32:21  <Qrntz> that's what I'm asking!
20:32:37  <Ptival> unless you're doing HPC...
20:40:15  <Ptival> anyway, given your constraints, you can use both
20:40:30  <Ptival> yes, you'll probably get better performance with hashtbl
20:40:56  <Ptival> but unless you're querying that structure like crazy, I doubt you'll be seeing much difference
20:41:38  <Qrntz> I fold it quite often
20:42:46  <Ptival> fold should be the same
20:42:49  <Qrntz> actually, I just discovered that I'm sorting the list I get by folding over that hash table in some places, so it would probably be better to use an ordered structure in the first place
20:46:01  <jonafan> have you profiled this code and shown that it's too slow?
20:47:38  <Qrntz> I profiled this code and the bottleneck isn't in this place
20:47:52  <Qrntz> I'm just obsessed with making everything «idiomatic» without losing performance
20:48:26  <jonafan> my preference is for persistant data structures like map
21:19:33  <orbitz> I want to have all of my library under a namespace of X, so a user can do X.Client or X.Server, for example.  Is best way to do that to make an X module and do module Server = X_server?
21:20:20  <pippijn> orbitz: you can make a pack
21:20:32  <orbitz> a pack with the devil?
21:20:50  <pippijn> ocamlc -pack or oasis Pack: true
21:21:34  <orbitz> God building in Ocaml is such a terrible experience
21:24:31  <adrien> ? packing is fairly simple
21:30:03  <orbitz> Lots of simple things makes one complex thing
21:31:53  <adrien> packing's main issue is that if you also access the functions from the same source repository as where you define them, it's possible that you access them without going through the pack
21:32:51  <pippijn> adrien: yes.. but only if you do, and also, why is that bad?
21:33:25  <adrien> it's not "bad", but it's doable while you might not want to do it
21:33:33  <pippijn> right
21:33:34  <pippijn> ok
21:33:46  <Qrntz> is there any way to «open» a module namespace for a dynamically loaded file from within the code that loads it?
21:34:02  <adrien> typically, you'd make a folder and pack everything in it with the idea that source files from outside that folder cannot skip the pack
21:34:35  <pippijn> I'm almost convinced that packs are a good thing, but not entirely, yet
21:51:33  <pippijn> I have a testcase here: http://paste.xinu.at/TOG/
21:52:12  <pippijn> the problem is that process_accept does not go past br2
21:52:17  <pippijn> Ssl.get_certificate hangs forever
21:52:43  <pippijn> well, I don't know that for sure, but it doesn't do anything
21:53:26  <pippijn> http://paste.xinu.at/7b7Pj/ <- server.pem
21:53:59  <pippijn> oh
21:54:05  <pippijn> so it doesn't hang
21:54:09  <pippijn> it produces an exception
21:54:31  <pippijn> so try_lwt ignores uncaught exceptions? that's not nice..
21:56:42  <pippijn> I don't understand.. if I re-raise it, I still see nothing
21:57:36  <pippijn> Printexc says it's an Ssl.Certificate_error.. is there any way to find out what was wrong with the cert?
21:58:15  <pippijn> maybe it's missing the CA
22:00:47  <mfp> pippijn: are you using Ssl.load_verify_locations somewhere?
22:03:52  <pippijn> mfp: is there any documentation for Ssl?
22:04:10  <mfp> AFAIK there's only the .mli
22:04:14  <pippijn> I'm looking at the mli and it has almost no comments
22:04:24  <pippijn> external load_verify_locations : context -> string -> string -> unit = "ocaml_ssl_ctx_load_verify_locations"
22:04:33  <pippijn> and ocaml is awesome like that.. it has no names in the declarations
22:04:47  <mfp> that's not what my .mli says
22:04:54  <pippijn> oh
22:05:03  <pippijn> it's the ml, sorry, I was looking in the wrong place
22:05:16  <pippijn> muuch better :)
22:05:21  <mfp> http://pastebin.com/4LhqZMTk
22:05:24  <pippijn> sorry
22:05:29  <mfp> ok :)
22:07:39  <mfp> you probably need to add Ssl.load_verify_locations ssl_context path_to_ca_cert ""   or such
22:08:03  <mfp> and something like   Ssl.set_verify ssl_context [Ssl.Verify_fail_if_no_peer_cert] None   to make sure there's a client certificate
22:08:17  <pippijn> mfp: I don't want that
22:08:23  <pippijn> the client certificate is optional
22:08:43  <pippijn> anyway, I tried the following things:
22:08:49  <mfp> then Verify_peer IIRC
22:09:00  <pippijn> hm, ok
22:11:21  <pippijn> ah
22:11:24  <pippijn> mfp: it works
22:11:42  <mfp> Qrntz: you cannot access the values defined in the dynamically loaded file directly, but you can have it register them somehow
22:12:29  <mfp> Qrntz: in particular, you can register a first-class module and then unpack it in the loader, giving you full access to whatever functionality you want to export in the loaded module :)
22:14:19  <pippijn> why isn't that exception propagated out?
22:15:01  <mfp> isn't it escaping Lwt_main.run?
22:15:37  <mfp> try_lwt automatically adds a  | exn -> raise_lwt exn   clause
22:16:06  <mfp> so it should be raising it
22:19:01  <mfp> pippijn: you might want to add a call to Ssl.verify for good measure (it'll raise if the client cert is not OK), even though in theory the handshake should be terminated according to SSL_CTX_set_verify(3) (it's not how that would be reported in the Ssl bindings)
22:19:19  <mfp> not *clear* how
22:20:12  <pippijn> set_verify does that, I think
22:20:17  <pippijn> and it does
22:20:23  <pippijn> in my code
22:20:35  <mfp> but there's only one  verify_callback, isn't there?
22:20:54  <mfp> one *value* of verify_callback, I mean
22:20:59  <pippijn> yeah
22:21:06  <pippijn> I set it to None
22:21:08  <pippijn> then it raised
22:21:33  <pippijn> http://paste.xinu.at/ybv/
22:23:59  <pippijn> mfp: I think <&> is the problem
22:24:21  <pippijn> a raised exception will wait at the infinite <&>
22:24:31  <mfp> what's that operator? I assume it was the pastebin garbling  >>
22:24:59  <mfp> ah Lwt.join
22:25:06  <pippijn> it's Lwt.join
22:25:18  <mfp> *assumed
22:25:22  <mfp> argh can't type tonight
22:26:29  <pippijn> I need to catch it, anyway, so I can close the socket
22:33:41  <Qrntz> mfp, hmm, that would indeed make the module accessible to the loaded ones, but I want to explicitly «open» it, as in let the loaded modules access the elements of it without using the module name
22:33:50  <Qrntz> (trying to mock up a plugin api)
22:34:46  <mfp> you want to use values defined in the "loader" module from the load*ed* one?
22:35:26  <Qrntz> well, yes, that would do it
22:35:39  <mfp> you should be able to do that directly
22:35:41  <Qrntz> (accessing without a qualified module name, though)
22:35:50  <Qrntz> hm
22:35:55  <mfp> just open Xxx then :)
22:37:35  <Qrntz> that's the problem — I don't want to «open» a module explicitly from within the loaded ones, that is, I need to access something from another module (preferrably, one specially meant for dynlinked modules) as if it were defined in the loaded module itself
22:38:01  <Qrntz> sort of like if it was defined in pervasives, I wouldn't go as far as augmenting pervasives though :-)
22:39:43  <mfp> I don't know any way to have that "implicit open" at load time
22:40:11  <mfp> would it be acceptable to preprocess the sources to prepend  open Mynewpervasives  at compile-time?
22:40:24  <Qrntz> that's what I just thought of, too
22:40:30  <Qrntz> perhaps
22:40:32  <mfp> (looks like a lot of work just to avoid  open X, but...)
22:40:42  <Qrntz> :-)
22:40:54  <mfp> batteries used to do that with camlp4
22:41:27  <Qrntz> maybe I'll think of something better, all I aim for is a plugin interface that's maximally transparent and «sandboxed» but is in ocaml
22:42:35  <Qrntz> the sandboxing part is achieved fairly easily with Dynlink.allow_only, now I'm trying to do something with transparency
22:43:09  <Qrntz> still, thanks for your help!
22:43:20  <mfp> Qrntz: does allow_only work for native code?
22:48:49  <Qrntz> good question
22:49:06  <Qrntz> it doesn't say anything about its byte/native compatibility in reference manual
22:50:03  <mfp> there's a note about allow_unsafe_modules being a NOP in native code, though
22:50:22  <mfp> so maybe allow_only does work
22:50:48  <Qrntz> hmm, yes, it seems to
22:51:00  <mfp> would like to know if you figure it out, but zzz time for me now
22:51:12  <Sablier> hi, does a lib like SDL-mixer who works for windows exist ?
22:51:17  * mfp afk
22:52:14  <Qrntz> mfp, on load it says «no implementation available» for the modules that were referenced but not allowed to be used, so yes, I presume it works
22:52:53  <pippijn> is there something that can parse /C=US/ST=State Name/L=Locality/O=Organisation/...?
22:52:55  <Qrntz> Sablier, ocamlsdl interfaces with sdl_mixer
22:53:07  <pippijn> and gives me a list of data?
22:53:44  <Sablier> yes of curse, but impossible to build ocaml-sdl :/
22:54:01  <Sablier> same for sfml
22:54:41  <Sablier> i try to solve that but i cant :s
22:56:06  <Sablier> and i dont see any compiled tools for that :/
23:14:42  <Qrntz> Sablier, it is certainly possible since I've succeeded in building it myself
23:14:47  <Qrntz> what exactly is your error
23:17:23  <Sablier> when i try make ocamlsdl i have this error
23:17:24  <Sablier> ocamlc.opt -ccopt "-I/mingw/include/SDL -D_GNU_SOURCE=1 -Dmain=SDL_main -g -O " sdl_stub.c
23:17:24  <Sablier> cc1.exe: erreur: unrecognized command line option '-mno-cygwin'
23:19:26  <Sablier> i use this tutorial : http://blogperso.univ-rennes1.fr/san.vu-ngoc/index.php/post/2011/02/20/Ocaml-and-SDL-on-windows
23:20:47  <Qrntz> oh, afraid I can't help you with windows setups
23:20:57  * Qrntz has no experience with that
23:21:01  <Sablier> i try to use cygwin after, but i have all the time errors dues to \r normes :p
23:21:05  <Sablier> oh np :p
23:21:14  <Qrntz> perhaps find where that option is defined and try to remove it?
23:22:14  <Sablier> I try, its not in the makefile of ocamlsdl
23:22:43  <Sablier> but i find 1 occurence of this flag in an ocaml makefile
23:22:50  <Sablier> i remove it but no change :'(
23:24:29  <Sablier> mmmh i think tomorrow i remove ocaml + cygwin + mingw + VC6 + all who can hurt that, and I retry :o
23:27:16  <Sablier> and i think i have to try with older versions of SDL & mingw when I see Last OCamL-SDL : Last modification date:	05-Nov-2002
23:30:51  <Sablier> mmmh thx Qrntz, i go to bed now, SDL is too much tiring :D
23:35:59  <Qrntz> np, Sablier, good night

Tiny iOS App in One Source File

Tue, 22 May 2012 19:00:00 +0000

May 22, 2012

Psi is a tiny, but complete, iOS test app in one source file, around 40 lines of Objective C. It’s nothing fancy, it just draws a Greek psi character on the screen. But I thought it might be useful to other folks. Sometimes it’s just nice to have a little example.

Psi can be very simple because it’s targeted at iOS 5, when Apple generalized the startup process to add storyboard support. A side effect is that it’s now possible to have no startup file (nib or storyboard file) at all. Psi shows how this can work. You can find a more detailed description of the iOS app launch sequence in a blog post by Ole Begemann, which is where I got the idea.

You can download the binary of Psi and run it directly in the iOS Simulator:

Psi Tiny iOS App Binary for iOS Simulator

To try out this version, unzip and double click on the launcher app PsiLauncher.app. You need to have Xcode and the iOS 5.0 Simulator installed for this to work. Psi appears on the second screen of apps—swipe to the left to see it. If it’s not there, you might need to change the version number to 5.1 in the Hardware -> Version menu.

Psi is so exceptionally simple that it’s really not very interesting to run—the figure above shows everything there is to see. However, it might be interesting to see how I packaged up an OS X app to run Psi in the simulator.

You can also download the sources and an Xcode project for building and running in the iOS Simulator or on an iOS device:

Psi Tiny iOS App Sources and Xcode project

To try out this version, unzip and double click on the Xcode project, Psi.xcodeproj.

You can also build and run from the command line using the runsim script if you are so inclined. After unzipping the sources, here’s what you need to do:

$ cd psi-1.0.0
$ make -f Makefile.iossim Psi
$ runsim Psi

For more about the runsim script, see Run iOS Simulator from the Command Line (Improved).

Finally, here is the full source of Psi.

/* psi.m     Tiny iOS App in one file
 *
 * (It just draws a Greek letter Psi.)
 *
 * Copyright (c) 2012 Psellos   http://psellos.com/
 * Licensed under the MIT License:
 *     http://www.opensource.org/licenses/mit-license.php
 */
#import <UIKit/UIKit.h>

@interface AppDelegate : NSObject <UIApplicationDelegate> {
}
@end

@implementation AppDelegate : NSObject
- (BOOL) application: (UIApplication *) anAppl
                didFinishLaunchingWithOptions: (NSDictionary *) opts
{
    UIViewController *vc = [[UIViewController alloc] init];
    UIWindow *w =
        [[UIWindow alloc] initWithFrame: [[UIScreen mainScreen] bounds]];
    UILabel *l = [[UILabel alloc] initWithFrame: [w bounds]];
    l.font = [UIFont fontWithName: @"MarkerFelt-Thin" size: 300.0];
    l.text = [NSString stringWithUTF8String: "\xce\xa8"];
    l.textColor =
        [UIColor colorWithRed: 0.141 green: 0.251 blue: 0.439 alpha: 1.0];
    l.textAlignment = UITextAlignmentCenter;
    l.backgroundColor =
        [UIColor colorWithRed: 0.800 green: 0.333 blue: 0.000 alpha: 1.0];
    vc.view = l;
    w.rootViewController = vc;
    [w makeKeyAndVisible];
    return YES;
}
@end

int main(int argc, char *argv[])
{
    NSAutoreleasePool * pool = [[NSAutoreleasePool alloc] init];
    int retVal = UIApplicationMain(argc, argv, nil, @"AppDelegate");
    [pool release];
    return retVal;
}

If you have any comments, corrections, or suggestions, leave them below or email me at jeffsco@psellos.com.

Posted by: Jeffrey

Caml Weekly News, 22 May 2012

Tue, 22 May 2012 12:00:00 +0000

Any library for reading/writing compressed files? / extunix 0.0.4, ocaml-mysql 1.1.1 and some more / Other Caml News

Iterating over the AST

virgile — Tue, 22 May 2012 09:45:00 +0000

Context

A facetious colleague who claims that he should be better writing his thesis but keeps committing Coq and OCaml files on Frama-C's repository, asked me the following question: Is there a function in Frama-C's kernel that can fold^[1] a function over all types that appear in a C program?

As it turns out, nothing is readily available for this exact purpose, but another facetious colleague could have come up with a terse answer: There's a visitor for that! Now, we can be a bit more helpful and actually write that visitor. Our end goal is to come up with a function

fold_typ: (Cil_types.typ -> 'a -> 'a) -> 'a -> 'a

so that fold_typ f init will be f t_1 (f t_2 (... (f t_n init)...)), where the t_i are all the C types appearing in a given C program (in no particular order).

Frama-C's visitors

The visitor pattern is a well-known object-oriented design pattern that can be used to perform a given action over all nodes of a complex data structure (in our case the Abstract Syntax Tree -AST for short- of a C program). Frama-C provides a generic visitor mechanism, built upon CIL's visitor, whose entry points can be found in the aptly named src/kernel/visitor.mli file. It is also documented in section 5.14 of the developer manual. In summary, you first define a class that inherits from the generic visitor and overrides the methods corresponding to the nodes you're interested in (in our case, this will be vtype for visiting uses of Cil_types.typ). Then you apply an object of this class to the function from the Visitor module that correspond to the subpart of the AST that you want to inspect (in our case, this will be the whole AST).

Basic version

The standard visitor does not provide anything to return a value outside of the AST. In fact, all the entry points in Visitor return a node of the same type that the node in which the visit starts (for visitors that don't perform code transformation, this is in fact physically the same node). But if you accept to sneak in a little bit of imperative code into your development -and by using Frama-C you've already accepted that- there is an easy way out: pass to the visitor a reference that it can update, and you just have to read the final value that reference is holding after the visit. The visitor then looks like the following:

class fold_typ_basic f acc =
object
  inherit Visitor.frama_c_inplace
  method vtype ty = acc:= f ty !acc; Cil.DoChildren
end

And that's it. Each time the visitor sees a type, it will apply f to ty and the result of the previous computations, stored in acc. fold_typ then just needs to call the visitor over the whole AST and give it a reference initialized with init:

let fold_typ f init =
  let racc = ref init in
  let vis = new fold_typ_basic f racc in
  Visitor.visitFramacFileSameGlobals vis (Ast.get());
  !racc

Don't do the same work twice

This first version, that is barely more than 10 LoC, works, but we can do a little better. Indeed, f will be called each time a type is encountered in the AST. In most cases, we want to call f once for any given type. This can be done quite simply by memoizing in an instance variable of our visitor the set of types encountered thus far. Frama-C's Cil_datatype module (cil/src/cil_datatype.mli) provides all the needed functions for that:

class fold_typ f acc =
object
  inherit Visitor.frama_c_inplace
  val mutable known_types = Cil_datatype.Typ.Set.empty
  method vtype ty =
    if Cil_datatype.Typ.Set.mem ty known_types then Cil.DoChildren
    else begin
      known_types <- Cil_datatype.Typ.Set.add ty known_types;
      acc:= f ty !acc;
      Cil.DoChildren
    end
end

Testing the infrastructure

It is now time to test if everything works smoothly. The following function will print the name and size of the type who has the biggest size in the analyzed program.

let test () =
  let f ty (maxty,maxsize as acc) =
    try
      let size = Cil.sizeOf_int ty in
      if size > maxsize then (ty,size) else acc
    with Cil.SizeOfError _ -> acc
  in
  let (ty,size) = fold_typ f (Cil.voidType,0) in
  Format.printf "Biggest type is %a,@ with size %d@." 
    !Ast_printer.d_type ty size

Since it is only a quick test, we don't do anything special if Cil complains that it cannot compute the size of a given type: we just stick to the maximal value computed so far.

File fold_typ.ml provides the code for the visitor and the test function. frama-c -load-script fold_typ.ml file.c should output something like

[kernel] preprocessing with "gcc -C -E -I.  file.c"
Biggest type is struct Ts,
with size 44

Exercises

How can you use the fold_typ class to define an iter_typ function that apply a function f returning unit to each type of the AST (val iter_typ: (Cil_types.typ -> unit) -> unit)?
Writing fold_typ is a bit overkill if you're going to apply it once in your plug-in. Write a specialized visitor that will do the same thing as the test function above.

Notes

[1] For those who are not familiar with functional programming: http://en.wikipedia.org/wiki/Fold_%...

0004934: ocamlbuild takes up to one minute to scan already done parts of Coq sources

Tue, 22 May 2012 09:26:12 +0000

Hi! I'm investigating the use of ocamlbuild for building Coq. I'd be glad to completely get rid of nasty Makefile stuff, but currently, two major issues prevent a complete switch to ocamlbuild. One is the lack of proper parallel compilation (see a forthcoming bug report). Another is that the delay needed by ocamlbuild to scan already built parts (> 1min here) prohibits using it for interactive development (edit a file, recompile, and so on). I've already had private discussions on this topic, without clear solutions, so I turn this into a proper bug report, for the records. Disclaimer: I clearly do not claim my myocamlbuild.ml file to be perfect, so this inefficiency can be caused by the way it is written. Any hints are welcome... Best regards, Pierre Letouzey Instructions to reproduce: 0) you'll need camlp5 installed 1) get a copy of coq development archive svn checkout <a href="svn://scm.gforge.inria.fr/svn/coq/trunk/">svn://scm.gforge.inria.fr/svn/coq/trunk/</a> [<a href="svn://scm.gforge.inria.fr/svn/coq/trunk/" target="_blank">^</a>] 2) ./configure -local -opt 3) Launch a first build via ocamlbuild (./build is a small wrapper) ./build 4) Launch ./build again Here, the first build says: Finished, 2716 targets (0 cached) in 00:13:10. While the second says: Finished, 2716 targets (2716 cached) in 00:01:07. With old-style make, the second run of make says instantly that nothing is to be done. NB: I've tried the patch for bug <a href="http://caml.inria.fr/mantis/view.php?id=4922">0004922</a>, unfortunately it doesn't help. We're not recompiling too many files here, simply taking too much time scanning cached things.

2012/05/21/Twitter

Mon, 21 May 2012 23:59:59 +0000

01:05:33  <raphscallion> ronald_duncan: :-) nw. I think functors are one of OCaml's biggest wins. Makes for very code flexible organisation.   | 204377356138328064
02:35:00  <levwalkin> Calling OCaml callback from C code invoked from OCaml runtime managed by Erlang. #layeredArchitecture #sarcasm   | 204399868523986944
10:24:09  <jonharrop> jaceklaskowski: HLVM is a VM I wrote using LLVM. http://t.co/yKta1Ljj   | 204517931231612929
10:43:16  <ontologiae> AtdGen #Ocaml serializer and deserializer of Ocaml data type in JSON or binary format   | 204522742991687681
10:47:54  <michael_pichon> "@ontologiae: AtdGen #Ocaml serializer and deserializer of Ocaml data type in JSON or binary format"   | 204523910857887744
13:51:07  <silentbicycle> pavlobaron: Basically, I wanted to get fluent again in a language complimentary to C, Lua, K, and Erlang. I'm too rusty at OCaml these days.   | 204570019336683521
14:50:11  <planet_ocaml> Makefile: typo in installoptopt entry: m Makefile m otherlibs/labltk/browser/Makefile.shared m otherlibs/labltk/... http://t.co/uUaVyp2d   | 204584883656466433
15:11:21  <mentby> Channels not closed on gc? #caml #ocaml http://t.co/PGv0tXe9   | 204590209592074240
23:15:29  <StackCodeReview> Is this binary search close to idiomatic OCaml? http://t.co/L50cOcmk #binarysearch   | 204712045264912386
23:31:16  <chaosape> In case you are looking for an OCaml style guide: http://t.co/7ROyNbhA   | 204716017472114688

2012/05/21/IRC

Mon, 21 May 2012 23:59:59 +0000

01:35:08  <wmeyer> thelema: now bootstraping from _oasis is possible, checkout the pull request
01:37:07  <wmeyer> thelema: sorry it's been taking that long - I sat down finally to odb today
07:22:21  <adrien> yarg
07:22:42  <adrien> most annoying thing when expanding lablgtk's API: not breaking API-compatibility for everyone =/
09:00:55  <ssbr_> Is there any function like this in the stdlib? fun a b ->(a, b)
09:05:18  <rixed> ssbr_: None that I know of. Intererstingly, there does not seams to have one in Batteries.Tuple2 neither :-(
09:06:08  <ssbr_> That is just weird. :(
09:06:56  <rixed> I think constructors are curryable in SML (bot not in OCaml), if this is what you are after. Although writting such a function would be trivial.
09:07:28  <ssbr_> rixed: nah, I'm building an association list from two lists
09:07:42  <ssbr_> List.map2 (fun a b -> (a,b)) list1 list2
09:07:47  <adrien>        val combine : 'a list -> 'b list -> ('a * 'b) list
09:08:28  <ssbr_> Oh.
09:08:29  <rixed> ssbr_: So List.map2 Cons list1 list2 may be valid SML (or close)
09:08:31  <ssbr_> Well that's convenient
09:09:07  <ssbr_> what kind of bot is adrien?
09:09:25  <ssbr_> it's a little weird to be able to pick the function that does what I want like that
09:09:36  <ssbr_> (List.combine is a thing that exists)
09:10:58  <adrien> sorry, forgot to mention that came from "man List" :-)
09:11:09  <ssbr_> adrien: what, you're human?
09:11:17  <ssbr_> maaaaaan
09:11:25  <ssbr_> you said it __right__ after I gave the function definition
09:11:29  <ssbr_> it was creepy.
09:11:45  <ssbr_> adrien: but thanks! :)
09:12:01  <adrien> I'd be more worrying if someone had managed to make a bot that reads IRC and properly gives such hints :-)
09:12:09  <Qrntz> a xavierbot with fuzzy logic
09:12:22  <adrien> (although we need a bette topic, an faq, and a bot which can lookup functions according to their types)
09:13:13  <ssbr_> well I was thinking how hard a problem it was. Figuring out which function in the stdlib is equivalent to a provided snippet of code is, well, undecidable to begin with
09:13:45  <rixed> ssbr_: actually, adrien is able to mimick a real person if you chat with him. But he is a bot obviously, you spotted it
09:14:24  <adrien> or a human who didn't have enough sleep and is waiting for some hardware to reboot =)
09:14:53  <flux> hah hah, how jolly and believeable nonsense those markov chains can generate!
09:14:58  <ssbr_> hum, data constructors aren't first class, huh?
09:15:05  <ssbr_> uhhh, why did I say "data"
09:15:05  <flux> ssbr_, sadly correct
09:15:40  <ssbr_> I have two constructors that mean the exact same thing except for one bit of extra information (logical negation). I do not like copy-pasting code like this. :<
09:15:51  <ssbr_> (I didn't make this design don't look at me.)
09:16:44  <flux> iirc there's a campl4 extension that achieves that trick
09:16:52  <flux> basically by generating functions for each constructor
09:17:14  <flux> well, it might not cover all cases of first-classfulness
09:17:53  <ssbr_> hyagh, and I can't even do (let a = (b, c) in SomeThing a) for SomeThing of a * b
09:18:21  <flux> use SomeThing of (a * b), they are succinctly different
09:18:28  <flux> of course, if that's what you already have, then never mind :)
09:18:48  <ssbr_> Why aren't all constructors nullary or unary?
09:18:59  <ssbr_> is there a benefit to SomeThing of a * b over SomeThing of (a * b) ?
09:19:13  <ssbr_> they aren't curried so there isn't that.
09:19:21  <flux> SomeThing of a * b has a more efficient memory representation, with some downsides as you noticed
09:23:07  <flux> with that the a and b are directly accompanied by the constructor, with the () form they are dereferenced through a pointer
09:23:46  <ssbr_> flux: I figured. Although I also figured that ocaml can certainly unbox things itself if it wants to. But I guess this is here so that the programmer can control it?
09:23:55  <ssbr_> except if it's here for programmer control, it sure is inconvenient
09:24:35  <rixed> ssbr_: OCamlc does not perform such complex things at code generation level.
09:25:04  <rixed> ssbr_: which is fine I think (predictability)
09:25:13  <flux> well, that kind of stuff affects the interfaces as well, so they cannot really be optimized. but, I suppose it could automatically do some conversion when extracting sub-patterns, although that would result in == perhaps not working in the expected way - but probably code relying on that would be broken on some level already anyway ;)
09:25:48  <ssbr_> rixed: well, it's only predictable if you read the type definition
09:26:00  <ssbr_> and you could always make it controllable by specifying whether you want the tuple to be boxed or unboxed
09:26:16  <ssbr_> (at said type definition).  I just don't like that the interface is so different between the two
09:26:38  <ssbr_> especially in ways that make my life harder :<
09:27:26  <rixed> ssbr_: I don't like this particular "syntax" neither. I'm just happy with a compiler that does not sometime inline the same data structures and sometime don't.
09:28:12  <flux> I just use always (a * b)
09:28:17  <flux> unless I have special performance needs
09:28:22  <ssbr_> flux: I think this is smart.
09:28:22  <flux> (haven't found such a situation yet)
09:28:51  <ssbr_> so sad that this isn't my code. :<
09:29:01  <rixed> flux: notice that if you have the requirement to deconstruct the type very fast then the ()ed form might be faster
09:29:10  <flux> rixed, well, true
14:38:54  <orbitz> Anyone know which godi works with Core?
14:38:59  <orbitz> Or at least I can install oasis with
14:41:16  <smondet> orbitz: Hi, last time I tried Core did not build with godi
14:41:37  <smondet> The core package in godi is quite outdated
14:41:42  <orbitz> Unfortunately
14:41:50  <orbitz> I'm trying to get godi to install oasis for me though
14:41:55  <orbitz> but that fails beause it depends o bin_prot?
14:42:33  <smondet> no, I think oasis depends only on type-conv (?)
14:42:56  <orbitz> Which it can't install :(
14:43:02  <orbitz> And installing oasis by hand looks annoing
14:43:12  <orbitz> Not terrible annoying, just the kind of annoying I don't feel like I should have to deal with in 2012
14:43:13  <smondet> you can get type-conv, core and more from that tar.gz: https://bitbucket.org/yminsky/ocaml-core/downloads
14:43:43  <orbitz> That download does not compile on mac
14:44:09  <orbitz> It's a vicious cycle
14:45:26  <smondet> I'e been there (but not on Mac): I got type-conv from godi, then compiled Oasis-0.3rc3 from sources, then got core from the bitbucket repo (and I had to put some code in comments)
14:47:05  <orbitz> Haha
14:47:07  <orbitz> Insanity
14:47:22  * orbitz shakes fist at Ocaml's poor build environment
14:48:01  <flux> careful with that fist or you might break the poor build environment even further :(
14:49:21  <orbitz> Hah
14:49:49  <orbitz> I wish we could replace the ocamlc ommunity with haskell since I want to write Ocaml and not Haskell but their build tools are awesome!
14:50:52  <avsm> the tarballs shouldnt need OASIS to be istnalled
14:51:27  <orbitz> For Core?
14:51:40  <orbitz> The proble mwith the Core tarball is it won't build for me, some 32bit 64bit missmatch error
15:03:55  <orbitz> I've taken to hacking godi database to fix the jane st deps
15:24:13  <hcarty> orbitz: If you need/want to install oasis you can give odb a try
15:24:47  <orbitz> hcarty: what is odb?
15:25:05  <hcarty> orbitz: https://github.com/thelema/odb
15:26:06  <orbitz> How many package systems are there or ocaml?
15:26:33  <hcarty> orbitz: Two?
15:26:43  <hcarty> orbitz: GODI and odb are two very different tools though
15:26:55  <orbitz> odb seems lik ea nice tool for bootstraping
15:27:00  <hcarty> GODI gets you up and running with OCaml + a set of packages
15:27:16  <orbitz> osx:tmp orbitz$ ocaml odb.ml oasis
15:27:16  <orbitz> Exception: Invalid_argument "String.sub".
15:27:17  <orbitz> :9
15:27:19  <orbitz> :(
15:27:19  <hcarty> ocamlbrew uses odb for bootstrapping an OCaml + base set of libraries
15:27:32  <hcarty> orbitz: What version of OCaml?
15:27:38  <orbitz> 3.12.0
15:27:41  <orbitz> Do I need latest?
15:28:57  <hcarty> orbitz: I don't think so?  But it's possible.  It works here for 3.12.1
15:29:48  <hcarty> orbitz: thelema may be able to help
15:29:50  <orbitz> Hrm, turning stacktrace on doesn't give me a stack trace...
15:30:06  <hcarty> The toplevel doesn't give traces
15:30:13  * orbitz shakes fist in rage yet again
15:30:24  <hcarty> Indeed
15:30:52  <hcarty> orbitz: Might I recommend Batteries? :-)
15:31:15  <orbitz> I want to play wtih Async :)
15:31:22  <hcarty> Lwt! :-)
15:31:31  <hcarty> I haven't tried Async so I can't compare the two
15:31:48  <orbitz> I've done Lwt, and like it, curious to see what Async is like
15:31:55  <orbitz> I have a feeling I won't like how Async handles errors
15:32:58  <orbitz> One advantage of Async apaprenly is the amount of C code is very small in it
15:52:22  <orbitz> hcarty: Hrm, same eror ron Ocaml 3.12.1
15:53:09  <hcarty> orbitz: Do you get the same error if you run odb.ml without any arguments?
15:53:20  <orbitz> Nope, i get a list of packages
15:53:36  <hcarty> orbitz: What is the command you are running?
15:53:45  <orbitz> ocaml odb.ml oasis
15:54:36  <hcarty> thelema: It looks like orbitz found a bug :-)
15:55:02  <hcarty> orbitz: I tested against the latest and got the same
15:55:13  <orbitz> Phew! Not jus tme
15:55:17  <hcarty> I must have had an outdated version of odb
15:55:31  <orbitz> I've got ot head to the airport thouhg, I'll check back tomorrow
15:55:53  <hcarty> orbitz: It will most likely be fixed by then
17:17:31  <pippijn> does ocsigen support SSL client authentication?
17:24:08  <pippijn> alternatively, does Lwt_ssl?
17:29:27  <thelema> hcarty: thanks for sending the bug report, I just noticed it, and it was a stupid mistake in testing whether a package name was a URI
17:29:40  <thelema> It should be fixed now.  Thanks orbitz as well
17:34:29  <hcarty> thelema: You're welcome - I'm glad it was a quick fix
17:36:51  <thelema> If only the jane street people would put their tarballs on some website, odb could easily install the new core
17:43:14  <wmeyer``> thelema: :)
17:43:31  <wmeyer``> thelema: I put the pull request "yesterday"
17:43:57  <thelema> wmeyer``: yes, I was looking through it.
17:44:28  <thelema> I hadn't gotten to where the buildtype detection was done
17:44:56  <thelema> ah, I see, the new function enclosure "try build using"
17:51:49  <wmeyer``> thelema: So it's a state machine - in simple words it tries first everything, and then when even make file files and the _oasis is present it tries to bootstrap by calling this function again with the Oasis_bootstrap, and the tries Oasis again - in fact in the begining I was thinking that we might want to try again everything, but the prefered order is setup.ml first anyway.
17:53:43  <wmeyer``> Maybe we should refactor how it's done currently in general the control flow in this function - I might have time to have a go today. Usually i think about monads at these times.
17:53:50  <smondet> pippijn: Once I got a Lwt_ssl server working with client certificates (not through ocsigen), I used example code from ocamlnet: http://docs.camlcity.org/docs/godipkg/3.12/godi-ocamlnet/doc/godi-ocamlnet/examples/rpc/queues/qclient_auth_ssl.ml
17:54:17  <thelema> wmeyer``: sorry, distracted by conference call, will chat more when it's done.
17:54:32  <wmeyer``> thelema: No problem, i am here whole evening.
18:03:41  <pippijn> smondet: that's a client
18:03:53  <pippijn> smondet: what does the server side look like?
18:04:25  <pippijn> ah, there is a qserver part
18:04:42  <pippijn> nice
18:04:46  <pippijn> smondet: thanks, that looks good
18:13:24  <smondet> pippijn: This part is the kind-of C.A.-independent part, you still need to check that certificates have not been revokated (and that depends on the CA implementation, I don't think it is standardized anywhere)
18:13:44  <pippijn> yeah
18:14:02  <pippijn> does ocaml have openssl ca bindings?
18:14:24  <wmeyer``> pippijn: I think I've seen some package doing that..
18:14:40  <wmeyer``> thelema: OK. I simplified a bit, added a commit to it
18:14:43  <pippijn> I could just call the openssl binary
18:15:06  <pippijn> but getting error codes from that is pretty crude
18:15:10  <pippijn> "failed" or "succeeded"
18:15:17  <smondet> I did not find any, I started trying to parse the files generated by openssl ca, but then I switched to create my own CA-like (I have a DB of users and roles already)
18:15:27  <pippijn> I'm not going to parse the console output from openssl
18:15:56  <pippijn> smondet: hm, maybe I'll do that..
18:16:24  <pippijn> the advantage of using openssl crl is that I can inspect it from outside with the openssl program
18:19:52  <smondet> pippijn: For inspection... I use sexplib.syntax + emacs :)
18:20:38  <pippijn> yeah
18:21:22  <pippijn> stupid debian..
18:21:35  <pippijn> their ocaml graphviz bindings need 3.11
18:21:38  <mrvn> ???
18:22:00  <pippijn> http://packages.debian.org/sid/libgv-ocaml
18:22:04  <mrvn> all ocaml packages need the ocaml they where build with.
18:22:13  <pippijn> I don't understand this
18:22:15  <mrvn> pippijn: file a bug for a binNMU.
18:22:29  <pippijn> oh, it doesn't exist for my architecture
18:22:45  <pippijn> it only exists in squeeze for my architecture
18:22:53  <pippijn> that's why..
18:22:54  <mrvn> The ocaml compiler produces object files that are incompatible between versions. So every time a new compiler version is uploaded all ocaml packages need to be recompiled.
18:23:15  <mrvn> pippijn: ahh, that explains why it wasn't rebuild then
18:23:17  <pippijn> mrvn: libgv-ocaml doesn't exist in testing
18:24:04  <mrvn> Check the bugs against ftp.debian.org, libgv-ocaml or the Removals.txt to see why it was removed.
18:24:17  <mrvn> or ask in debian-ocaml or on the ML.
18:25:19  <pippijn> [auto-cruft] NBS (no longer built by graphviz)
18:26:26  <pippijn> Debian Bug report logs: Bugs in package libgv-ocaml
18:26:29  <pippijn> No reports found!
18:26:44  <mrvn> so check the graphviz source why it stoped building it.
18:27:07  <pippijn> it needs --enable-ocaml
18:27:20  <pippijn> as far as I can tell, it didn't stop
18:27:34  <pippijn> there is even a libgv-ocaml.install in the debian directory
18:51:32  <thelema> wmeyer``: simpler is better, looking at new commit now
18:52:04  <mrvn> less code - less bugs. :)
18:52:37  <thelema> mrvn: yes, I was sad when I crossed the 500LoC boundary on odb.ml
18:53:31  <mrvn>   * Remove ocaml bindings (Closes: #647435)
18:54:57  <mrvn> pippijn: Seems to have been removed for lack of use. So if you have a use case just ask for it to be added back.
18:55:49  <mrvn> and add a fix for the bug
18:57:03  <thelema> wmeyer``: I think I like the simpler code better.  What about if oasis isn't installed - auto-installing it?
19:08:07  <mfp> argh why does epoll_wait specify the timeout in milliseconds
19:08:46  <mfp> and why does Lwt_unix.sleep 0.0001 sleep for >3(!)ms   http://pastebin.com/Y7k7nEGx
19:13:14  <flux> when ever does sleep X say that you are not going to sleep >X?-)
19:13:58  <flux> mfp, strace -tt or -ttt might be nice
19:14:19  <flux> too bad it doesn't (?) log the time spent in a system call
19:14:48  <flux> ah, -T!
19:15:19  <adrien> strace does everything ;-)
19:15:38  <flux> except handle threads properly
19:15:40  <mfp> ah yes    epoll_wait(3, {}, 64, 1)                = 0 <0.003857>
19:16:32  <flux> what are you doing where you need 1000 us sleeps?
19:16:52  <adrien> Sleep(1);
19:17:17  <flux> mfp, btw, what happens if you sleep for 0.002 s?
19:19:02  <mrvn> flesk: then you wake up some time after 0.002s have passed
19:19:26  <mrvn> flux: ^^^
19:20:12  <mfp> I have a lwt thread which writes to disk (and can fsync), which is awaken when there's dirty data. But I want to amortize the write/fsync cost by doing group commits, and thus not Lwt.wakeup (actually, Lwt_condition.broadcast) right away, but after say 0.1ms (small delay so as to avoid adding too much latency in the non-fsync, non-concurrent case)
19:21:30  <mrvn> I hope you don't delay the write too
19:21:52  <mfp> ?
19:22:18  <mrvn> if you write and then wait before the fsync then the FS has a chance to already write the data
19:22:35  <mfp> flux: Lwt_unix.sleep 0.002 gives these times  0.0058191 0.0039639 0.0039780 0.0040021 0.0041711 0.0038090 0.0039890 0.0039980 0.0039930
19:23:26  <mfp> mrvn: I don't control the fsync time myself; I'm writing to a leveldb DB and setting a fsync:bool param on write.
19:25:36  <mfp> I guess I could write with fsync:false and then perform a dummy fsynced write, so they'll go to the same append-only file and flush at once, but I don't really see the point (unless I want to allow non-fsynced commits, but that's another story)
19:26:21  <mrvn> that would just write the data twice.
19:28:07  <mfp> would it? I mean the dummy fsynced write would not carry the same data (just some dummy record)
19:28:47  <mrvn> it would still need to write at least one block. The point of doing the fsync later would be so that the fsync wouldn't block
19:28:48  <mfp> btw. Thread.delay 0.0001 gives a very constant  0.0001581  delay
19:29:57  <mfp> the fsync is done in a separate thread, not the main one with the Lwt event loop
19:30:20  <mfp> so no blocking
19:34:27  <mfp> select(0, [], [], [], {0, 10})          = 0 (Timeout) <0.000079>    so there's no pb sleeping for ~100us
19:44:05  <flux> ah, I realized that epoll_wait cannot do shorter sleeps than 1 millisecond
19:44:18  <flux> new cool interface for the win?
19:44:52  <flux> someone somewhere decided that a new interface should only do sleeps >= 1000 us :)
19:45:20  <flux> well, you could probably use setitimer to get shorter sleeps
19:46:39  <flux> mfp, maybe you can choose to use some other async backend
19:47:12  <mfp> yeah, could switch to the one based on select
19:49:03  <mfp> but can also keep the libev one and synthesize a faster Lwt_unix.sleep by sleeping in a new thread and sending a notification; that would add some 20+20us latency or so
19:50:19  * mfp afk
19:50:43  <adrien> aren't you overengineering this a bit?
19:53:01  <mrvn> That's a bit like if (time < 1000) { for i = 0; i < time * 3438; ++i) { } } else { usleep(time); }
20:19:56  <bitbckt> cd
20:19:59  <bitbckt> oops :)
20:20:02  <bitbckt> >_> <_<
20:20:21  <jonafan> LOL NOOB!!!!!!
20:20:34  <jonafan> let's keep it on topic bitbckt
20:20:43  <bitbckt> cd ~/src/ocaml?
20:20:48  <jonafan> better
20:20:49  <bitbckt> :P
21:37:26  <pippijn> oasis supports packs
21:37:30  <pippijn> that's nice
22:30:04  <eikke> is there any way to create a 'string' using alloca() in ocaml? :P
22:30:54  <pippijn> that would be a bad idea
22:30:58  <pippijn> and no
22:31:15  <pippijn> why do you want that?
22:32:25  <pippijn> I have a question about Lwt_ssl: http://paste.xinu.at/617/
22:32:46  <eikke> some code in which I would create a small (like, 16-byte) buffer which will never escape local scope, but will be mutated in the algorithm
22:32:58  <pippijn> when I make a connection, it starts using 100% CPU time
22:33:31  <pippijn> eikke: okay.. in ocaml, you don't allocate a lot of things on the stack
22:33:41  <eikke> I'm aware of that
22:34:06  <eikke> we're (more as a challenge) trying to squeeze the last overhead out of some very tight loop
22:34:14  <eikke> in a couple of languages, comparing generated asm,...
22:34:23  <pippijn> ah
22:34:43  <pippijn> well, if you don't mind the challenge, try storing the bytes of that string in a few ints
22:34:55  <eikke> ocaml is pretty good, but calling camlBuffer__add_char is painful (I wish that could be inlined :D)
22:35:11  <pippijn> uh
22:35:27  <pippijn> before you start using Buffer, try just using a string
22:35:42  <pippijn> if you want performance
22:36:04  <pippijn> String.make 16 ' '
22:37:11  <pippijn> it won't be on the stack, but that probably doesn't matter too much
22:37:29  <pippijn> ocaml's runtime is pretty good with locality of reference
22:39:44  <smondet> pippijn: I think Lwt_ssl.accept needs to ba called on a socket already accepted by Lwt_unix.accept (I think I found readable server-side code in ocsigenserver's darcs)
22:39:58  <pippijn> oh
22:40:59  <smondet> http://ocsigen.org/darcsweb/?r=ocsigenserver.dev;a=headblob;f=/src/server/ocsigen_server.ml#l976
22:41:11  <smondet> they use Lwt_unix.accept_n
22:41:55  <pippijn> yeah, that's better
22:42:21  <pippijn> it works
22:55:58  <pippijn> smondet: can I get the Ssl.socket from an Lwt_ssl.socket?
22:56:38  <pippijn> I want to get the client certificate
22:57:19  <smondet> Ah yes I had that problem :)
22:57:37  <pippijn> I thought so
22:57:41  <pippijn> that's why I sask you :)
22:57:45  <smondet> So, it's not possible, in current version, I reported it and they fixed it
22:57:55  <smondet> (very quickly)
22:57:59  <pippijn> oh
22:58:03  <pippijn> so I have to wait?
22:58:11  <pippijn> or get HEAD/trunk/tip/whatever
22:58:43  <smondet> in the maintime my code was using Obj.magic (my code has not been touched in a few months, I had other stuf to do)
22:59:05  <pippijn> oh, it segfaulted with that
22:59:59  <pippijn> how did you do that?
23:00:07  <pippijn> I tried Ssl.get_certificate (List.hd (Obj.magic socket))
23:00:47  <pippijn> bad luck, it's not the first field
23:01:04  <smondet>  pippijn: http://pastebin.com/aEtYBJwX
23:01:56  <pippijn> oh
23:02:05  <pippijn> ok, so it's the second field
23:11:30  <pippijn> well, it doesn't work :)
23:11:39  <pippijn> it hangs, doing.. I don't know
23:12:29  <pippijn> ok, I see
23:12:45  <pippijn> Ssl. Certificate_error
23:19:02  <pippijn> Ssl.get_certificate hangs?
23:21:46  <wmeyer``> thelema: We could do it
23:22:04  <pippijn> http://paste.xinu.at/m4Eoj/
23:22:05  <wmeyer``> thelema: I think it should be highly sort of optional to install oasis
23:22:30  <wmeyer``> thelema: the problem is that we don't want developers just to rely on this feature - right?
23:23:33  <wmeyer``> thelema: I would think that oasis might be the first thing to install - but do you do it with autotools? btw however it's the only day way you can do with the source control.
23:31:54  <eikke> pippijn: fwiw, using a string increases runtime with >100% and more-than-doubles GC rate in a microbench ;)
23:32:16  <pippijn> eikke: vs. using a Buffer?
23:32:30  <eikke> sorta, yes
23:32:37  <eikke> it's already usign a buffer
23:33:10  <pippijn> how about using a global string?
23:33:30  <pippijn> or a semi-global one
23:34:57  <eikke> thought about that as well, but I always think that feels really ugly and not really concurrency-safe
23:35:29  <pippijn> you can pass the string in
23:35:32  <pippijn> as a context
23:35:56  <eikke> that pushes allocation cost to the caller, wont help. + I might need to enlarge the string on-the-go
23:36:48  <pippijn> can you pre-calculate the maximum string size?
23:37:09  <pippijn> maybe have a conservative estimate
23:43:17  <eikke> pippijn: sure I could. fwiw, here's the thing: https://gist.github.com/2765432
23:43:47  <pippijn> eikke: I'm going to sleep now
23:43:49  <pippijn> I'll look at it tomorrow
23:44:20  <eikke> sure, nn

0005127: wrong location after directive

Mon, 21 May 2012 21:30:45 +0000

The locations in the Camlp4 ast are completely wrong for the material that follows a directive. For a two line file #use "y.ml";; let a = 5 the location in the StVal node for line 2 has start_line = 3 (should be 2) start_bol = 1 (should be 14) start_off = 1 (should be 14) the stop location is also wrong.

RegStab 2.0

Mon, 21 May 2012 15:45:38 +0000

RegSTAB is a SAT-solver able to deal with formula schemas: you can give it a scheme of formulas such as "/\i=1..n P_i -> P_i+1" (where n is a variable) and it will be able to answer you if *all the formulas of this form (i.e. for every value of n) are unsatisfiable*, i.e. it can treat at once an infinite set of propositional formulas.

ocaml-winsvc 0.0.1

Mon, 21 May 2012 15:29:37 +0000

The library to make OCaml program act as a Windows service.