Changements intéressants: * la configuration upstream d'une branche évolue: --set-upstream est dépréciée, et est remplacée par --set-upstream-to=reference. Au delà du changement de nom, c'est la syntaxe qui évolue pour être plus simple et plus logique. En effet, avec --set-upstream, vous deviez taper:

git branch --set-upstream my-branch origin/dev

--set-upstream est un flag, qui précise que le refspec de base de la branche créée est aussi l'upstream. Ce n'est pas évident, et l'on aurait tendance à interpréter la commande dans l'autre sens, dans la tradition des commandes Unix. Avec la nouvelle syntaxe, on tape:

git branch --set-upstream-to=origin/dev my-branch

Cette syntaxe est bien plus compréhensible. L'ancienne syntaxe est dépréciée. Si vous ne connaissez toujours pas les magies et bienfaits des branches upstream, je vous invite à (re)lire ce billet.

• git cherry-pick respecte désormais l'ordre des commits tels que passés en arguments. Précédemment, il rejouait les commits suivant la chronologie de commit.

• git blame est désormais plus précis lors des merge : il reconnait les conflits résolus comme venant d'un des commits parents plutôt que simplement étant des modifications non committées (sha1=0000000)

N'hésitez pas comme d'habitude à commenter ce billet si vous voyez des changements intéressants.

Changelog

• A credential helper for Win32 to allow access to the keychain of the logged-in user has been added.

• An initial port to HP NonStop.

• A credential helper to allow access to the Gnome keyring has been added.

• When "git am" sanitizes the "Subject:" line, we strip the prefix from "Re: subject" and also from a less common "re: subject", but left the even less common "RE: subject" intact. Now we strip that too.

• It was tempting to say "git branch --set-upstream origin/master", but that tells Git to arrange the local branch "origin/master" to integrate with the currently checked out branch, which is highly unlikely what the user meant. The option is deprecated; use the new "--set-upstream-to" (with a short-and-sweet "-u") option instead.

• "git cherry-pick" learned the "--allow-empty-message" option to allow it to replay a commit without any log message.

• After "git cherry-pick -s" gave control back to the user asking help to resolve conflicts, concluding "git commit" used to need to be run with "-s" if the user wants to sign it off; now the command leaves the sign-off line in the log template.

• "git daemon" learned the "--access-hook" option to allow an external command to decline service based on the client address, repository path, etc.

• "git difftool --dir-diff" learned to use symbolic links to prepare a temporary copy of the working tree when available.

• "git grep" learned to use a non-standard pattern type by default if a configuration variable tells it to.

• Accumulated updates to "git gui" has been merged.

• "git log -g" learned the "--grep-reflog=" option to limit its output to commits with a reflog message that matches the given pattern.

• "git merge-base" learned the "--is-ancestor A B" option to tell if A is an ancestor of B. The result is indicated by its exit status code.

• "git mergetool" now allows users to override the actual command used with the mergetool.$name.cmd configuration variable even for built-in mergetool backends.

• "git rebase -i" learned the "--edit-todo" option to open an editor to edit the instruction sheet.

• "git svn" has been updated to work with SVN 1.7.

• "git p4" learned the "--conflicts" option to specify what to do when encountering a conflict during "p4 submit".

• Git ships with a fall-back regexp implementation for platforms with buggy regexp library, but it was easy for people to keep using their platform regexp by mistake. A new test has been added to check this.

• The "check-docs" build target has been updated and greatly simplified.

• The test suite is run under MALLOC_CHECK_ when running with a glibc that supports the feature.

• The documentation in the TeXinfo format was using indented output for materials meant to be examples that are better typeset in monospace.

• Compatibility wrapper around some mkdir(2) implementations that reject parameters with trailing slash has been introduced.

• Compatibility wrapper for systems that lack usable setitimer() has been added.

• The option parsing of "git checkout" had error checking, dwim and defaulting missing options, all mixed in the code, and issuing an appropriate error message with useful context was getting harder. The code has been reorganized to allow giving a proper diagnosis when the user says "git checkout -b -t foo bar" (e.g. "-t" is not a good name for a branch).

• Many internal uses of a "git merge-base" equivalent were only to see if one commit fast-forwards to the other, which did not need the full set of merge bases to be computed. They have been updated to use less expensive checks.

• The heuristics to detect and silently convert latin1 to utf8 when we were told to use utf-8 in the log message has been transplanted from "mailinfo" to "commit" and "commit-tree".

• Messages given by "git -h" from many subcommands have been marked for translation.

• The attribute system may be asked for a path that itself or its leading directories no longer exists in the working tree, and it is fine if we cannot open .gitattribute file in such a case. Failure to open per-directory .gitattributes with error status other than ENOENT and ENOTDIR should be diagnosed, but it wasn't.

• When looking for $HOME/.gitconfig etc., it is OK if we cannot read them because they do not exist, but we did not diagnose existing files that we cannot read.

• When "git am" is fed an input that has multiple "Content-type: ..." header, it did not grok charset= attribute correctly.

• "git am" mishandled a patch attached as application/octet-stream (e.g. not text/*); Content-Transfer-Encoding (e.g. base64) was not honored correctly.

• "git blame MAKEFILE" run in a history that has "Makefile" but not "MAKEFILE" should say "No such file MAKEFILE in HEAD", but got confused on a case insensitive filesystem and failed to do so.

• Even during a conflicted merge, "git blame $path" always meant to blame uncommitted changes to the "working tree" version; make it more useful by showing cleanly merged parts as coming from the other branch that is being merged.

• It was unclear in the documentation for "git blame" that it is unnecessary for users to use the "--follow" option.

• Output from "git branch -v" contains "(no branch)" that could be localized, but the code to align it along with the names of branches was counting in bytes, not in display columns.

• "git cherry-pick A C B" used to replay changes in A and then B and then C if these three commits had committer timestamps in that order, which is not what the user who said "A C B" naturally expects.

• A repository created with "git clone --single" had its fetch refspecs set up just like a clone without "--single", leading the subsequent "git fetch" to slurp all the other branches, defeating the whole point of specifying "only this branch".

• Documentation talked about "first line of commit log" when it meant the title of the commit. The description was clarified by defining how the title is decided and rewording the casual mention of "first line" to "title".

• "git cvsimport" did not thoroughly cleanse tag names that it inferred from the names of the tags it obtained from CVS, which caused "git tag" to barf and stop the import in the middle.

• Earlier we made the diffstat summary line that shows the number of lines added/deleted localizable, but it was found irritating having to see them in various languages on a list whose discussion language is English, and this change has been reverted.

• "git fetch --all", when passed "--no-tags", did not honor the "--no-tags" option while fetching from individual remotes (the same issue existed with "--tags", but the combination "--all --tags" makes much less sense than "--all --no-tags").

• "git fetch" over http had an old workaround for an unlikely server misconfiguration; it turns out that this hurts debuggability of the configuration in general, and has been reverted.

• "git fetch" over http advertised that it supports "deflate", which is much less common, and did not advertise the more common "gzip" on its Accept-Encoding header.

• "git fetch" over the dumb-http revision walker could segfault when curl's multi interface was used.

• "git gc --auto" notified the user that auto-packing has triggered even under the "--quiet" option.

• After "gitk" showed the contents of a tag, neither "Reread references" nor "Reload" updated what is shown as the contents of it when the user overwrote the tag with "git tag -f".

• "git log --all-match --grep=A --grep=B" ought to show commits that mention both A and B, but when these three options are used with --author or --committer, it showed commits that mention either A or B (or both) instead.

• The "-Xours" backend option to "git merge -s recursive" was ignored for binary files.

• "git p4", when "--use-client-spec" and "--detect-branches" are used together, misdetected branches.

• "git receive-pack" (the counterpart to "git push") did not give progress output while processing objects it received to the puser when run over the smart-http protocol.

• When you misspell the command name you give to the "exec" action in the "git rebase -i" instruction sheet you were told that 'rebase' is not a git subcommand from "git rebase --continue".

• The subcommand in "git remote" to remove a defined remote was "rm" and the command did not take a fully-spelled "remove".

• The interactive prompt that "git send-email" gives was error prone. It asked "What e-mail address do you want to use?" with the address it guessed (correctly) the user would want to use in its prompt, tempting the user to say "y". But the response was taken as "No, please use 'y' as the e-mail address instead", which is most certainly not what the user meant.

• "git show --format='%ci'" did not give the timestamp correctly for commits created without human readable name on the "committer" line.

• "git show --quiet" ought to be a synonym for "git show -s", but wasn't.

• "git submodule frotz" was not diagnosed as "frotz" being an unknown subcommand to "git submodule"; the user instead got a complaint that "git submodule status" was run with an unknown path "frotz".

• "git status" honored the ignore=dirty settings in .gitmodules but "git commit" didn't.

• "gitweb" did not give the correct committer timezone in its feed output due to a typo.

La présentation d'une heure s'appelle «Pourquoi Git ?», destinée aux gens pressés. Puis le lendemain deux ateliers «Initiation» et «Allez plus loin».

7h de Git au programme, si vous aviez envie de vous y mettre, c'est l'occasion !

Néanmoins, je n'ai plus autant de temps à consacrer à gitfr, notamment à cause de mes nouveaux projets et cela se ressens sur le peu de billet sur ce blog. Alors, si vous êtes partant pour écrire voir reprendre tout en partie des activités de gitfr, c'est avec plaisir que je transmettrais le flambeau.

Si cela vous intéresse : seb@gitfr.net.

Note : le coût est de 15€ / an pour le nom de domaine que je prends à ma charge, le reste étant hébergé sur GitHub. Le site peut être revue de fond en comble si besoin.

• Une option --root à rebase pour modifier aussi le commit root.

• Une option -x à rebase pour éxécuter une commande à chaque application d'un patch.

• On peut visualiser la documentation en HTML avec git help -w $cmd

• git status est plus précis quand le dépôt est dans un état intermédiaire (lors d'un merge, un rebase, etc).

N'hésitez pas comme d'habitude à commenter ce billet si vous voyez des changements intéressants.

Le changelog

• Git can be told to normalize pathnames it read from readdir(3) and all arguments it got from the command line into precomposed UTF-8 (assuming that they come as decomposed UTF-8), in order to work around issues on Mac OS.

I think there still are other places that need conversion (e.g. paths that are read from stdin for some commands), but this should be a good first step in the right direction.

• Per-user $HOME/.gitconfig file can optionally be stored in $HOME/.config/git/config instead, which is in line with XDG.

• The value of core.attributesfile and core.excludesfile default to $HOME/.config/git/attributes and $HOME/.config/git/ignore respectively when these files exist.

• Logic to disambiguate abbreviated object names have been taught to take advantage of object types that are expected in the context, e.g. XXXXXX in the "git describe" output v1.2.3-gXXXXXX must be a commit object, not a blob nor a tree. This will help us prolong the lifetime of abbreviated object names.

• "git apply" learned to wiggle the base version and perform three-way merge when a patch does not exactly apply to the version you have.

• Scripted Porcelain writers now have access to the credential API via the "git credential" plumbing command.

• "git help" used to always default to "man" format even on platforms where "man" viewer is not widely available.

• "git clone --local $path" started its life as an experiment to optionally use link/copy when cloning a repository on the disk, but we didn't deprecate it after we made the option a no-op to always use the optimization. The command learned "--no-local" option to turn this off, as a more explicit alternative over use of file:// URL.

• "git fetch" and friends used to say "remote side hung up unexpectedly" when they failed to get response they expect from the other side, but one common reason why they don't get expected response is that the remote repository does not exist or cannot be read. The error message in this case was updated to give better hints to the user.

• "git help -w $cmd" can show HTML version of documentation for "git-$cmd" by setting help.htmlpath to somewhere other than the default location where the build procedure installs them locally; the variable can even point at a http:// URL.

• "git rebase [-i] --root $tip" can now be used to rewrite all the history leading to "$tip" down to the root commit.

• "git rebase -i" learned "-x " to insert "exec " after each commit in the resulting history.

• "git status" gives finer classification to various states of paths in conflicted state and offer advice messages in its output.

• "git submodule" learned to deal with nested submodule structure where a module is contained within a module whose origin is specified as a relative URL to its superproject's origin.

• A rather heavy-ish "git completion" script has been split to create a separate "git prompting" script, to help lazy-autoloading of the completion part while making prompting part always available.

• "gitweb" pays attention to various forms of credits that are similar to "Signed-off-by:" lines in the commit objects and highlights them accordingly.

• "mediawiki" remote helper (in contrib/) learned to handle file attachments.

• "git p4" now uses "Jobs:" and "p4 move" when appropriate.

• vcs-svn has been updated to clean-up compilation, lift 32-bit limitations, etc.

• Some tests showed false failures caused by a bug in ecryptofs.

• We no longer use AsciiDoc7 syntax in our documentation and favor a more modern style.

• "git am --rebasing" codepath was taught to grab authorship, log message and the patch text directly out of existing commits. This will help rebasing commits that have confusing "diff" output in their log messages.

• "git index-pack" and "git pack-objects" use streaming API to read from the object store to avoid having to hold a large blob object in-core while they are doing their thing.

• Code to match paths with exclude patterns learned to avoid calling fnmatch() by comparing fixed leading substring literally when possible.

• "git log -n 1 -- rarely-touched-path" was spending unnecessary cycles after showing the first change to find the next one, only to discard it.

• "git svn" got a large-looking code reorganization at the last minute before the code freeze.

• "git submodule add" was confused when the superproject did not have its repository in its usual place in the working tree and GIT_DIR and GIT_WORK_TREE was used to access it.

• "git commit --amend" let the user edit the log message and then died when the human-readable committer name was given insufficiently by getpwent(3).

• Ne pas demander explicitivement de voir ma présentation. Moralité pratiquement personne ne l'avait vu et donc la plupart avaient des lacunes théoriques importantes.

• Ne pas définir un niveau (débutant, intermédiaire, confirmé), j'avais donc des parfaits débutants comme des utilisateurs avec plusiueurs mois d'expérience.

J'ai donc opté pour un atelier avec des gros morceaux de présentation donnés au fil de l'eau quand cela était nécessaire pour comprendre les commandes a taper.

Je fais donc le constat suivant :

• Je donnerais un atelier si et seulement si j'ai donné une présentation juste avant. Non obligatoire mais me permet de ne pas revenir sur la théorie (et tant pis pour ceux qui estiment de ne pas en avoir besoin mais sont à la ramasse).

• Fixer un niveau pour l'atelier.

Cela peut donc donner une journée complète de Git (pres + atelier débutant + atelier intermédiaire). Mais avec plus de 20 présentations et 7 ateliers, cela ne me fait pas vraiment peur. J'ai d'ailleurs totalement improvisé cet atelier / pres de 3h :).

• Fini la configuration par défaut qui pousse toutes les branches connues sur le serveur. C'est maintenant le mode simple qui prend le relais, mode qui ne pousse (si et seulement si elle possède un nom identique sur le serveur) que le branche courante. C'est une configuration bien plus pertinente pour les débutants. Néanmoins je vous préconise le mode upstream comme dit dans ce billet.

• Un outil en shell du nom de git-subtree.sh est ajoutée dans la section contrib, outil qui simplifie le découpage ou la fusion de dépôt. Cela permet par exemple de se dispenser de submodules en intégrant un dépôt dans un autre dépôt.

N'hésitez pas comme d'habitude à commenter ce billet si vous voyez des changements intéressants.

Le changelog

• A new mode for push, "simple", which is a cross between "current" and "upstream", has been introduced. "git push" without any refspec will push the current branch out to the same name at the remote repository only when it is set to track the branch with the same name over there. The plan is to make this mode the new default value when push.default is not configured.

• A couple of commands learned the "--column" option to produce columnar output.

• A third-party tool "git subtree" is distributed in contrib/

• A remote helper that acts as a proxy and caches ssl session for the https:// transport is added to the contrib/ area.

• Error messages given when @{u} is used for a branch without its upstream configured have been clarified.

• Even with the "-q"uiet option, "checkout" used to report setting up tracking. Also "branch" learned the "-q"uiet option to squelch informational message.

• Your build platform may support hardlinks but you may prefer not to use them, e.g. when installing to DESTDIR to make a tarball and untarring on a filesystem that has poor support for hardlinks. There is a Makefile option NO_INSTALL_HARDLINKS for you.

• The smart-http backend used to always override GIT_COMMITTER_* variables with REMOTE_USER and REMOTE_ADDR, but these variables are now preserved when set.

• "git am" learned the "--include" option, which is an opposite of existing the "--exclude" option.

• When "git am -3" needs to fall back to an application of the patch to a synthesized preimage followed by a 3-way merge, the paths that needed such treatment are now reported to the end user, so that the result in them can be eyeballed with extra care.

• The output from "diff/log --stat" used to always allocate 4 columns to show the number of modified lines, but not anymore.

• "git difftool" learned the "--dir-diff" option to spawn external diff tools that can compare two directory hierarchies at a time after populating two temporary directories, instead of running an instance of the external tool once per a file pair.

• The "fmt-merge-msg" command learned to list the primary contributors involved in the side topic you are merging in a comment in the merge commit template.

• "git rebase" learned to optionally keep commits that do not introduce any change in the original history.

• "git push --recurse-submodules" learned to optionally look into the histories of submodules bound to the superproject and push them out.

• A 'snapshot' request to "gitweb" honors If-Modified-Since: header, based on the commit date.

• "gitweb" learned to highlight the patch it outputs even more.

• "git svn" used to die with unwanted SIGPIPE when talking with an HTTP server that uses keep-alive.

• "git svn" learned to use platform specific authentication providers, e.g. gnome-keyring, kwallet, etc.

• "git p4" has been moved out of the contrib/ area and has seen more work on importing labels as tags from (and exporting tags as labels to) p4.

• Bash completion script (in contrib/) have been cleaned up to make future work on it simpler.

• An experimental "version 4" format of the index file has been introduced to reduce on-disk footprint and I/O overhead.

• "git archive" learned to produce its output without reading the blob object it writes out in memory in its entirety.

• "git index-pack" that runs when fetching or pushing objects to complete the packfile on the receiving end learned to use multiple threads to do its job when available.

• The code to compute hash values for lines used by the internal diff engine was optimized on little-endian machines, using the same trick the kernel folks came up with.

• "git apply" had some memory leaks plugged.

• Setting up a revision traversal with many starting points was inefficient as these were placed in a date-order priority queue one-by-one. Now they are collected in the queue unordered first, and sorted immediately before getting used.

• More lower-level commands learned to use the streaming API to read from the object store without keeping everything in core.

• The weighting parameters to suggestion command name typo have been tweaked, so that "git tags" will suggest "tag?" and not "stage?".

• Because "sh" on the user's PATH may be utterly broken on some systems, run-command API now uses SHELL_PATH, not /bin/sh, when spawning an external command (not applicable to Windows port).

• The API to iterate over the refs/ hierarchy has been tweaked to allow walking only a subset of it more efficiently.

• "git submodule init" used to report "registered for path ..." even for submodules that were registered earlier. (cherry-pick c1c259e jl/submodule-report-new-path-once later to maint).

• "git diff --stat" used to fully count a binary file with modified execution bits whose contents is unmodified, which was not quite right.

• gitk a pas mal de petites évolutions.

• Un éditeur s'ouvre lors d'un git-commit pour pousser les gens à expliquer les changements.

• On peut cloner qu'un répertoire.

N'hésitez pas comme d'habitude à commenter ce billet si vous voyez des changements intéressants dans votre façon de travailler avec Git.

Le changelog

• various "gitk" updates.

• Teams for localizing the messages from the Porcelain layer of commands are starting to form, thanks to Jiang Xin who volunteered to be the localization coordinator. Translated messages for simplified Chinese, Swedish and Portuguese are available.

• The configuration mechanism learned an "include" facility; an assignment to the include.path pseudo-variable causes the named file to be included in-place when Git looks up configuration variables.

• A content filter (clean/smudge) used to be just a way to make the recorded contents "more useful", and allowed to fail; a filter can now optionally be marked as "required".

• Options whose names begin with "--no-" (e.g. the "--no-verify" option of the "git commit" command) can be negated by omitting "no-" from its name, e.g. "git commit --verify".

• "git am" learned to pass "-b" option to underlying "git mailinfo", so that a bracketed string other than "PATCH" at the beginning can be kept.

• "git clone" learned "--single-branch" option to limit cloning to a single branch (surprise!); tags that do not point into the history of the branch are not fetched.

• "git clone" learned to detach the HEAD in the resulting repository when the user specifies a tag with "--branch" (e.g., "--branch=v1.0"). Clone also learned to print the usual "detached HEAD" advice in such a case, similar to "git checkout v1.0".

• When showing a patch while ignoring whitespace changes, the context lines are taken from the postimage, in order to make it easier to view the output.

• "git diff --stat" learned to adjust the width of the output on wider terminals, and give more columns to pathnames as needed.

• "diff-highlight" filter (in contrib/) was updated to produce more aesthetically pleasing output.

• "fsck" learned "--no-dangling" option to omit dangling object information.

• "git log -G" and "git log -S" learned to pay attention to the "-i" option. With "-i", "log -G" ignores the case when finding patch hunks that introduce or remove a string that matches the given pattern. Similarly with "-i", "log -S" ignores the case when finding the commit the given block of text appears or disappears from the file.

• "git merge" in an interactive session learned to spawn the editor by default to let the user edit the auto-generated merge message, to encourage people to explain their merges better. Legacy scripts can export GIT_MERGE_AUTOEDIT=no to retain the historical behavior. Both "git merge" and "git pull" can be given --no-edit from the command line to accept the auto-generated merge message.

• The advice message given when the user didn't give enough clue on what to merge to "git pull" and "git merge" has been updated to be more concise and easier to understand.

• "git push" learned the "--prune" option, similar to "git fetch".

• The whole directory that houses a top-level superproject managed by "git submodule" can be moved to another place.

• "git symbolic-ref" learned the "--short" option to abbreviate the refname it shows unambiguously.

• "git tag --list" can be given "--points-at