Git v2.2 Release Notes ====================== Updates since v2.1 ------------------ Ports * Building on older MacOS X systems automatically sets the necessary NO_APPLE_COMMON_CRYPTO build-time option. UI, Workflows & Features * "git config --edit --global" starts from a skeletal per-user configuration file contents, instead of a total blank, when the user does not already have any. This immediately reduces the need for a later "Have you forgotten setting core.user?" and we can add more to the template as we gain more experience. * "git stash list -p" used to be almost always a no-op because each stash entry is represented as a merge commit. It learned to show the difference between the base commit version and the working tree version, which is in line with what "git show" gives. * Sometimes users want to report a bug they experience on their repository, but they are not at liberty to share the contents of the repository. "fast-export" was taught an "--anonymize" option to replace blob contents, names of people and paths and log messages with bland and simple strings to help them. * "log --date=iso" uses a slight variant of ISO 8601 format that is made more human readable. A new "--date=iso-strict" option gives datetime output that is more strictly conformant. * A broken reimplementation of Git could write an invalid index that records both stage #0 and higher stage entries for the same path. We now notice and reject such an index, as there is no sensible fallback (we do not know if the broken tool wanted to resolve and forgot to remove higher stage entries, or if it wanted to unresolve and forgot to remove the stage#0 entry). * The "pre-receive" and "post-receive" hooks are no longer required to consume their input fully (not following this requirement used to result in intermittent errors in "git push"). * The pretty-format specifier "%d", which expanded to " (tagname)" for a tagged commit, gained a cousin "%D" that just gives the "tagname" without frills. Performance, Internal Implementation, etc. * The API to manipulate the "refs" is currently undergoing a revamp to make it more transactional, with the eventual goal to allow all-or-none atomic updates and migrating the storage to something other than the traditional filesystem based one (e.g. databases). * We no longer attempt to keep track of individual dependencies to the header files in the build procedure, relying on automated dependency generation support from modern compilers. * In tests, we have been using NOT_{MINGW,CYGWIN} test prerequisites long before negated prerequisites e.g. !MINGW were invented. The former has been converted to the latter to avoid confusion. * Looking up remotes configuration in a repository with very many remotes defined has been optimized. * There are cases where you lock and open to write a file, close it to show the updated contents to external processes, and then have to update the file again while still holding the lock, but the lockfile API lacked support for such an access pattern. * The API to allocate the structure to keep track of commit decoration has been updated to make it less cumbersome to use. * An in-core caching layer to let us avoid reading the same configuration files number of times has been added. A few commands have been converted to use this subsystem. * Various code paths have been cleaned up and simplified by using "strbuf", "starts_with()", and "skip_prefix()" APIs more. * A few codepaths that died when large blobs that would not fit in core are involved in their operation have been taught to punt instead, by e.g. marking too large a blob as not to be diffed. * A few more code paths in "commit" and "checkout" have been taught to repopulate the cache-tree in the index, to help speed up later "write-tree" (used in "commit") and "diff-index --cached" (used in "status"). * A common programming mistake to assign the same short option name to two separate options is detected by parse_options() API to help developers. * The code path to write out the packed-refs file has been optimized, which especially matters in a repository with a large number of refs. * The check to see if a ref $F can be created by making sure no existing ref has $F/ as its prefix has been optimized, which especially matters in a repository with a large number of existing refs. * "git fsck" was taught to check contents of tag objects a bit more. * "git hash-object" was taught a "--literally" option to help debugging. Also contains various documentation updates and code clean-ups. Fixes since v2.1 ---------------- Unless otherwise noted, all the fixes since v2.1 in the maintenance track are contained in this release (see the maintenance releases' notes for details). * "git log --pretty/format=" with an empty format string did not mean the more obvious "No output whatsoever" but "Use default format", which was counterintuitive. * Implementations of "tar" that do not understand an extended pax header would extract the contents of it in a regular file; make sure the permission bits of this file follows the same tar.umask configuration setting. * "git -c section.var command" and "git -c section.var= command" should pass the configuration differently (the former should be a boolean true, the latter should be an empty string). * Applying a patch not generated by Git in a subdirectory used to check the whitespace breakage using the attributes for incorrect paths. Also whitespace checks were performed even for paths excluded via "git apply --exclude=" mechanism. * "git bundle create" with date-range specification were meant to exclude tags outside the range, but it didn't. * "git add x" where x that used to be a directory has become a symbolic link to a directory misbehaved. * The prompt script checked $GIT_DIR/ref/stash file to see if there is a stash, which was a no-no. * Pack-protocol documentation had a minor typo. * "git checkout -m" did not switch to another branch while carrying the local changes forward when a path was deleted from the index. * With sufficiently long refnames, "git fast-import" could have overflown an on-stack buffer. * After "pack-refs --prune" packed refs at the top-level, it failed to prune them. * Progress output from "git gc --auto" was visible in "git fetch -q". * We used to pass -1000 to poll(2), expecting it to also mean "no timeout", which should be spelled as -1. * "git rebase" documentation was unclear that it is required to specify on what the rebase is to be done when telling it to first check out . (merge 95c6826 so/rebase-doc later to maint). * "git push" over HTTP transport had an artificial limit on number of refs that can be pushed imposed by the command line length. (merge 26be19b jk/send-pack-many-refspecs later to maint). * When receiving an invalid pack stream that records the same object twice, multiple threads got confused due to a race. (merge ab791dd jk/index-pack-threading-races later to maint). * An attempt to remove the entire tree in the "git fast-import" input stream caused it to misbehave. (merge 2668d69 mb/fast-import-delete-root later to maint). * Reachability check (used in "git prune" and friends) did not add a detached HEAD as a starting point to traverse objects still in use. (merge c40fdd0 mk/reachable-protect-detached-head later to maint). * "git config --add section.var val" used to lose existing section.var whose value was an empty string. (merge c1063be ta/config-add-to-empty-or-true-fix later to maint). * "git fsck" failed to report that it found corrupt objects via its exit status in some cases. (merge 30d1038 jk/fsck-exit-code-fix later to maint). * Use of "--verbose" option used to break "git branch --merged". (merge 12994dd jk/maint-branch-verbose-merged later to maint). * Some MUAs mangled a line in a message that begins with "From " to ">From " when writing to a mailbox file and feeding such an input to "git am" used to lose such a line. (merge 85de86a jk/mbox-from-line later to maint). * "rev-parse --verify --quiet $name" is meant to quietly exit with a non-zero status when $name is not a valid object name, but still gave error messages in some cases.