Skip to main content

Browse the archive

Revision be66a6c43dcba42c56f66a8706721a76098f8e25 authored by Johannes Schindelin on 25 April 2009, 09:57:14 UTC, committed by Junio C Hamano on 25 April 2009, 16:49:21 UTC 
Add an option not to use link(src, dest) && unlink(src) when that is unreliable
 
It seems that accessing NTFS partitions with ufsd (at least on my EeePC)
has an unnerving bug: if you link() a file and unlink() it right away,
the target of the link() will have the correct size, but consist of NULs.

It seems as if the calls are simply not serialized correctly, as single-stepping
through the function move_temp_to_file() works flawlessly.

As ufsd is "Commertial software" (sic!), I cannot fix it, and have to work
around it in Git.

At the same time, it seems that this fixes msysGit issues 222 and 229 to
assume that Windows cannot handle link() && unlink().

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Acked-by: Johannes Sixt <j6t@kdbg.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
1 parent 785a985
Raw File
RelNotes-1.5.4.5.txt 
GIT v1.5.4.5 Release Notes
==========================

Fixes since v1.5.4.4
--------------------

 * "git fetch there" when the URL information came from the Cogito style
   branches/there file did not update refs/heads/there (regression in
   1.5.4).

 * Bogus refspec configuration such as "remote.there.fetch = =" were not
   detected as errors (regression in 1.5.4).

 * You couldn't specify a custom editor whose path contains a whitespace
   via GIT_EDITOR (and core.editor).

 * The subdirectory filter to "git filter-branch" mishandled a history
   where the subdirectory becomes empty and then later becomes non-empty.

 * "git shortlog" gave an empty line if the original commit message was
   malformed (e.g. a botched import from foreign SCM).  Now it finds the
   first non-empty line and uses it for better information.

 * When the user fails to give a revision parameter to "git svn", an error
   from the Perl interpreter was issued because the script lacked proper
   error checking.

 * After "git rebase" stopped due to conflicts, if the user played with
   "git reset" and friends, "git rebase --abort" failed to go back to the
   correct commit.

 * Additional work trees prepared with git-new-workdir (in contrib/) did
   not share git-svn metadata directory .git/svn with the original.

 * "git-merge-recursive" did not mark addition of the same path with
   different filemodes correctly as a conflict.

 * "gitweb" gave malformed URL when pathinfo stype paths are in use.

 * "-n" stands for "--no-tags" again for "git fetch".

 * "git format-patch" did not detect the need to add 8-bit MIME header
   when the user used format.header configuration.

 * "rev~" revision specifier used to mean "rev", which was inconsistent
   with how "rev^" worked.  Now "rev~" is the same as "rev~1" (hence it
   also is the same as "rev^1"), and "rev~0" is the same as "rev^0"
   (i.e. it has to be a commit).

 * "git quiltimport" did not grok empty lines, lines in "file -pNNN"
   format to specify the prefix levels and lines with trailing comments.

 * "git rebase -m" triggered pre-commit verification, which made
   "rebase --continue" impossible.

As usual, it also comes with many documentation fixes and clarifications.
back to top