On Feb 6, 2007, at 11:55, Frodak wrote:
> I think the source of the issue is how cygwin's
> version of stat performs exe magic. So if file exists
> it'll get return information about file, if file
> doesn't exist but file.exe does exist, stat returns
> information about file.exe.
>
> If I recall correctly most things that NEED to work
> around this have been updated to use something other
> then stat to see if the object exists.
>
> I don't think you'll run into the problem often
> though. I think that the directory would normally
> always be exported first.
>
> Sorry, I couldn't be more helpful.
No, that's quite helpful. I would say it confirms that we have an
undesired behavior in Subversion. The order of items being exported
or checked out is rather random, as far as I can tell, so anyone with
the problematic repository contents (foo.exe and foo in same
directory) could run into the same problem.
I would say that Subversion should be updated to not use stat, or if
that's not possible, then it should stat foo, and if it thinks foo
exists, then it should also stat foo.exe, and if that also exists,
then it should compare those two in as accurate a way as possible
(can we look at inode numbers or their equivalent?) to determine if
what it thinks is foo is really foo.exe. And... if those checks can
be made to only occur on Windows under cygwin then I guess that would
be best.
Is Subversion calling stat directly, or is it using an APR function
for that? If an APR function, then these fixes should be made in APR,
not Subversion, right?
--
To reply to the mailing list, please use your mailer's Reply To All
function
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org
Received on Tue Feb 6 22:58:20 2007