Hello,

/usr/tmp is not (used) on OpenBSD.

We have specific patch to do this, but I prefer to have this modification upstream.

2.5: http://www.openbsd.org/cgi-bin/cvsweb/ports/lang/python/2.5/patches/patch-Lib_tempfile_py
2.7: http://www.openbsd.org/cgi-bin/cvsweb/ports/lang/python/2.7/patches/patch-Lib_tempfile_py
3.2: http://www.openbsd.org/cgi-bin/cvsweb/ports/lang/python3/3.2/patches/patch-Lib_tempfile_py

Is it possible to add a specific location to remove /usr/tmp of the dirlist?

Attached file is the patch to do this:
--- Lib/tempfile.py.orig        Thu Aug  4 17:05:24 2011
+++ Lib/tempfile.py     Tue Aug 23 12:55:40 2011
@@ -138,6 +138,8 @@
     # Failing that, try OS-specific locations.
     if _os.name == 'nt':
         dirlist.extend([ r'c:\temp', r'c:\tmp', r'\temp', r'\tmp' ])
+    elif _sys.platform.startswith('openbsd'):
+        dirlist.extend([ '/tmp', '/var/tmp' ])
     else:
         dirlist.extend([ '/tmp', '/var/tmp', '/usr/tmp' ])
, '/var/tmp' ])

Thanks,
Remi.

> /usr/tmp is not (used) on OpenBSD.

How is that a problem?
Those are just possible temporary directories to try, so if it doesn't exist, it won't be used (typically ENOENT will be catched when opening a tmpfile).
Also, since it's listed last, that's not a performance problem either.

Because it's not exactly a bug, can you only apply the patch to Python 3.3?

> can you only apply the patch ...

Hum, I mean can *we* only apply ...

> Because it's not exactly a bug, can you only apply the patch to Python 3.3?

Well, we could, but:
- it doesn't fix a bug
- it doesn't bring any obvious benefit (performance, security, etc)
- it introduces an OS-specific case for no reason

I'm open to changing my mind though, maybe I'm missing something.

I was reporting this because I thought it was better to include this upstream.
However, I understand if you don't want to include this specific case.

We can keep this patch in our tree, I understand your position.

> We can keep this patch in our tree, I understand your position.

I could be missing something, but why do you have this patch in the first place?
If it turns out to be useful, then it's of course better to include it upstream, I just don't see what this patch is supposed to solve/improve.

> I could be missing something, but why do you have this patch in the first place?

I'm recently working on porting Python for OpenBSD, and this patch exists since python2.5 (since 4 years in our tree). So I don't have all responses, but I think this issue could be closed/rejected, and it seems that this patch could be safely removed from our port tree.

Thanks for all information.

Remi.

> it seems that this patch could be safely removed from our port tree.

Alright, I'm closing then.

> Thanks for all information.

And thanks for your help and bug reports concerning OpenBSD-specific issues!