Re: [PATCH] Severe database corruption when running without standard file descriptors on unix

On Fri, 2004-09-03 at 11:35, Tobias Ringström wrote:
> problem with the patch in that it's using #ifndef WIN32.

I don't really consider this a problem. Feature testing is the right
thing when dealing with non-portable Unix features (because otherwise
you get into "AIX version 5 has this, but version 4 doesn't, and every
few years a new version of Unix comes along and you have to update the
tests"), but isn't so much the right thing when dealing with portable
Unix features which don't apply to Windows.

It can (and has been) argued that "#ifndef WIN32" is a dumb test for
Unix, but we use it all over the code base. If and when we settle on a
better preprocessor test for it, we can do a global replacement, but
that's a larger issue than any particular bit of Unix-specific code.

The patch looks fine to me. I don't think it's really an app's
responsibility to cope when invoked incorrectly like this, but it's
cheap and it obviously would have helped at least one person avoid
shooting himself in the foot.

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Fri Sep 3 18:05:29 2004