Re: [RFC PATCH] cygwin: disallow backslashes in file names

[Johannes Schindelin <Johannes.Schindelin@gmx.de>]

[-- Attachment #1: Type: text/plain, Size: 4474 bytes --]

Hi Adam,

On Tue, 27 Apr 2021, Adam Dinwoodie wrote:

> On Mon, 26 Apr 2021 at 20:56, Adam Dinwoodie wrote:
> >
> > On Mon, 26 Apr 2021 at 15:08, Johannes Schindelin wrote:
> > >
> > > Hi Adam,
> > >
> > > On Sat, 24 Apr 2021, Adam Dinwoodie wrote:
> > > > Notes:
> > > >     The patch to read-cache.c is the one I've applied downstream as the Cygwin Git
> > > >     maintainer to resolve this vulnerability, and I've manually tested that it
> > > >     resolves the vulnerability, so that's the change I'd recommend anyone who needs
> > > >     to build Git on Cygwin themselves take until there's something officially in
> > > >     the Git source code.
> > > >
> > > >     I'm much less convinced by my approach for the test script.  I definitely think
> > > >     it's worth having a test here, but the test as written still fails, as the test
> > > >     seems to be looking for the error message "directory not empty", but running
> > > >     the test on Cygwin produces the error "cannot create submodule directory d\a".
> > > >     I'm not sure why that difference exists, and whether the correct approach would
> > > >     be to (a) ensure the error messages are consistent across platforms or (b) to
> > > >     change the test to expect the appropriate error on the appropriate platform.
> > >
> > > Wasn't there something in Cygwin that _allowed_ backslashes as file name
> > > characters? I vaguely remember that the ASCII characters forbidden by
> > > Windows were mapped into some "private page".
> > >
> > > Maybe that is responsible for the difference here?
> >
> > So there is special handling of a bunch of characters like ":" that
> > are valid as parts of filenames on most *nix systems, but which aren't
> > valid on Windows, by substituting them for characters in the Unicode
> > "private use area" space. Backslash isn't one of those characters,
> > though; quoting
> > https://cygwin.com/cygwin-ug-net/using-specialnames.html (which I just
> > checked myself to be sure): "The backslash has to be exempt from this
> > conversion, because Cygwin accepts Win32 filenames including
> > backslashes as path separators on input."
> >
> > Which is not to say this special handling _isn't_ the cause of the
> > difference here, but it's not so simple as that. If nobody spots an
> > explanation I've missed, I'll start digging into the code and strace
> > to work out exactly what's causing the difference in behaviour.
>
> I've worked out what's going wrong here: the "prevent git~1 squatting
> on Windows" test is actually testing a selection of different Windows
> path oddities, which are handled differently between Git for Windows
> and Cygwin Git. The specific behaviour here is the handling of a
> directory called "d."; Git for Windows (I assume in the MSYS2 layer)
> follows the standard Windows convention of treating "d." and "d" as
> identical filenames, while Cygwin sticks to its general design
> philosophy of mostly emulating *nix systems, allowing objects with
> both filenames to exist in the same directory (and causing pain for
> most non-Cygwin applications that try to interact with them).
>
> Essentially this test is checking a bunch of different oddities about
> path handling on Windows. Some things – such as handling backslashes –
> are common to both Cygwin and MSYS2; some – such as handling trailing
> periods – aren't. So I expect the solution here will be to have
> separate tests for (a) Git for Windows, (b) Cygwin Git, and (c) common
> behaviour.

Ah, that would explain things. Thank you so much for digging!

Ciao,
Dscho

>
> > > >     I'm also not convinced by my approach of adding a "WINDOWS" prerequisite to
> > > >     test-lib.sh. I went with this as I couldn't immediately see a way to pass
> > > >     prerequisites on an "any" rather than "all" basis to test_expect_success, and
> > > >     this would allow us to simplify all the tests that currently have
> > > >     "!MINGW,!CYGWIN" as prerequisites, but it still feels a bit clunky to me.
> > >
> > > Right, the only way I could think of it would be
> > >
> > >         test_lazy_prereq 'test_have_prereq MINGW || test_have_prereq CYGWIN'
> > >
> > > Your approach looks fine to me, though.
> >
> > Grand, okay. I'll stick with that for now, then, and follow up with a
> > patch to tidy up the other prerequisites at some point in the future.
> >
> > Adam
>