Re: Karl, your r3484 chante to io.c:svn_io_copy_file...

Karl Fogel wrote:

>(Of course, you would *think* that if you call apr_get_file_info() on
>Platform Foo, and pass the results along as perms to apr_file_copy
>also on Platform Foo, that this could not be a problem. Silly me.)
>
Ah, but this is not Platform Foo we're talking about -- it's Platform
Eeeek! :-)
Seriously: apr_file_info_get() does try hard to map the Windows ACLs to
Unix permissions, but the mapping is simply not possible in general.

>>For now, I'll disable the copy_perms flag on Windows, so that I can run
>>the test, and put that warning back, so that people don't forget about
>>the issue. We can decide about a real solution after we've discussed
>>this a bit more.
>>
>>
>
>I think the cleanest solution would be to make the tmp copy with perms
>(except on Win32) and then rely on the rename behavior described
>above.
>
>How does that sound to you?
>
>
That would bring the implementation close to what it was, except for the
additional rename, correct? And the warning put back. Yes, I agree.
Later on we can teach apr_file_copy to honour APR_FILE_SOURCE_PERMS on
Windows, too.

-- 
Brane Čibej   <brane_at_xbc.nu>   http://www.xbc.nu/brane/
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org