[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[bug-diffutils] bug#64811: Win-version now broken
From: |
Paul Eggert |
Subject: |
[bug-diffutils] bug#64811: Win-version now broken |
Date: |
Tue, 25 Jul 2023 11:20:19 -0700 |
User-agent: |
Mozilla/5.0 (X11; Linux x86_64; rv:102.0) Gecko/20100101 Thunderbird/102.13.0 |
On 2023-07-25 08:07, Gisle Vanem wrote:
Tried it, still the same error.
I suggest looking into it with a debugger.
with a trailing '\', e.g. this fails:
diff.exe -r -u3 -Hb %CD% %CD%\Git-latest\
with:
diff.exe: F:\MinGW32\src\gnu\GNU-diff\Git-latest\: No such file or
directory
Here are calls that diff executes for that on my platform. (getdents is
a low-level call that implements readdir.) You can try running under a
debugger and seeing what happens on your platform.
fstatat(AT_FDCWD, "%CD%", ..., 0) = 0
fstatat(AT_FDCWD, "%CD%\\Git-latest\\", ..., 0) = 0
openat(AT_FDCWD, "%CD%", O_RDONLY|O_DIRECTORY) = 3
fstat(3, ...) = 0
getdents(3, ...) = 2624
getdents(3, ...) = 0
openat(AT_FDCWD, "%CD%\\Git-latest\\", O_RDONLY|O_DIRECTORY) = 4
fstat(4, ...) = 0
getdents(4, ...) = 1488
getdents(4, ...) = 0
Not sure if the attachment of my 'GnuDiff.diff' comes through.
Otherwise it's here:
http://www.watt-32.net/misc/gnudiff-diff-before.jpg
I didn't see it, and I'd rather not deal with images of text: they're
inconvenient, sometimes hard to read, and (not saying you're doing this
of course) occasionally contain hidden attempts to break into the
developer's environment and for this reason often go missing or unread.
Please just send text.
I also have to state, that merging these patches by hand (since
I've done a lot of private patches to 'src/*.c'), is very hard
I'm hoping that whatever further changes are needed for MS-Windows are
relatively minor, so that you don't have to maintain your own set of
patches by hand, whatever they are. I'm also hoping any such changes are
localized to Gnulib files so that I don't have to worry about them as a
diffutils maintainer and so that they can be shared by other GNU
programs with minimal hassle. The idea is that the MS-Windows platform
is lower priority for GNU and that concerns about it shouldn't need to
hamper mainline GNU development.
With that in mind, you might want to focus on the Gnulib test failure
first, as they should be smaller and easier to debug.