If double clicked on from explorer it will start a new command prompt but
invoking the script from a current command prompt just adds the ssh keys
to the prompt.
Fix for issue #128 where the prompt causes the console to beep if there are
non-ASCII characters in the path. This patch replaces all non-ASCII characters
with a question mark when expanding the prompt
Signed-off-by: Anselm Kruis <a.kruis@science-computing.de>
Signed-off-by: Pat Thoyts <patthoyts@users.sourceforge.net>
Somehow git rebase -i seems to ignore GIT_EDITOR and only uses
GIT_SEQUENCE_EDITOR. Set both variables to not introduce a regression.
Signed-off-by: Stefan Naewe <stefan.naewe@gmail.com>
Sometimes people commit all kinds of stuff before performing a merging
rebase, and that unfortunately also entails merges that look like '-s
ours' type merges. Therefore, it is not good enough to look whether
the merge looks like that, but we also have to ask the user if that's
the start of the latest merging rebase.
Only when we find that the commit subject contains 'merging-rebase',
we trust that we have found the correct merge.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
It was discussed recently that it is rather confusing to keep that
script when we retired it in favor of merging-rebase.sh...
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
These two files have been pulled in from CoreUtils and were built against
an old set of libraries. We do not ship these binaries so we can avoid
rebuilding against the current libintl-8 library although this should
be done sometime. Used when running the test suite.
Signed-off-by: Pat Thoyts <patthoyts@users.sourceforge.net>
As discussed in issue #113 it is useful to have a method for converting
between unicode and utf-8 for writing smudge/clean filters. To assist in
this we shall include a functional iconv.exe in the release.
A side effect is we now also ship libintl-8.dll.
Signed-off-by: Pat Thoyts <patthoyts@users.sourceforge.net>
In case the user has modified GIT_SSH manually after the last
installation, prefer that path over the one from the previous
installation for the Plink location in case it points to a valid
Plink executable.
Signed-off-by: Sebastian Schuberth <sschuberth@gmail.com>
Instead of limiting the accepted Plink location to a directory with
certain executables, allow any executable that has "plink" in its name.
Signed-off-by: Sebastian Schuberth <sschuberth@gmail.com>
Piping the result of the `ls` command (alias) to a file will contain escape characters because the `--color` setting for `ls` is defaulting to `--color=always` instead of `--color=auto` (auto means, it will only produce color control codes on a real tty)
Erik noted that the new compilation options cause libcurl to be linked to
zlib1.dll which we do not include. Removing the ZLIB_PATH parameter
keeps it statically linked.
The winssl code can fail to operate behind a proxy. It tries to check for
certificate revocation, fails and then aborts the ssl negotiations.
Therefore reverting back to openssl for ssl support.
Signed-off-by: Pat Thoyts <patthoyts@users.sourceforge.net>
On Windows XP ccache crashed when used in compiling git. Applying the patch
from ccache-win32 issue #8 resolved this problem and a number of other
patches from this tracker have been included.
Signed-off-by: Pat Thoyts <patthoyts@users.sourceforge.net>
Currently, if client (i.e. share/WinGit/portable-release.sh) asks
copy-files.sh not to remove git builtins, we just keep them in bin/ and
libexec/git-core/ . The problem is, later, when those executables are
stripped, though initially they were identical, they all become
different, because:
For PE, strip (and everything from binutils) puts current time into
produced object in PE header into TimeDateStamp [1]:
---- 8< ---- bfd/peXXigen.c
H_PUT_32 (abfd, time (0), filehdr_out->f_timdat);
and that hurts, if later a user wants to hardlink'ify portable git
install.
Though strip has --preserve-dates, as said above, it does not affect the
_contents_ of the generated object file - only mtime/ctime are
preserved.
Let's preserve builtins ourselves - we already keep list of them in
etc/fileList-builtins.txt for msysgit installer (it relinks them at
install time), and for portable version, lets first remove all builtins
except git.exe, then strip, and then restore builtins bit-to-bit equal
to git.exe . This way, later hardlinkifying will work.
[1] http://www.csn.ul.ie/~caolan/publink/winresdump/winresdump/doc/pefile2.html
Signed-off-by: Kirill Smelkov <kirr@mns.spb.ru>
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
In effect, this reverts the change of the escape sequence from
\033]0;
to
\033[0m
due to an incorrect understanding of their purposes in 4e4e3f3.
Conflicts:
etc/profile
In 4e4e3f3, the 0; and bell etc. were incorrectly interpreted as having
to do with colors. In fact, it is responsible for setting the window
title. [1]
[1] http://www.faqs.org/docs/Linux-mini/Xterm-Title.html#ss3.1
Signed-off-by: Tay Ray Chuan <rctay89@gmail.com>
This is in line with
http://msdn.microsoft.com/en-us/library/windows/desktop/aa511450.aspx
saying
"Don't abuse the desktop - keep users in control. If your target users
are likely to use your program frequently, provide an option during setup
to put a shortcut on the desktop, unselected by default."
Signed-off-by: Sebastian Schuberth <sschuberth@gmail.com>
Just make that component part of no types. In the same run change
documentation and code a bit to reflect the new idea that components that
should be installed by default in a fresh installation must have the
"default" type assigned.
Signed-off-by: Sebastian Schuberth <sschuberth@gmail.com>
Sebastian Schuberth
Matt Clarkson
Pat Thoyts
Anselm Kruis
Johannes Schindelin
Stefan Naewe
Karsten Blees
Erik Faye-Lund
Bernd
Kirill Smelkov
Tay Ray Chuan