At 8:16 PM -0800 3/24/00, Prymmer/Kahn wrote:
>This is just a quick note to let folks know that 5.6.0
>final release did build OK on that strange Alpha that
>had both a volnam of USER and an overriding rooted
>logical name USER. Although I had suspected that I would
>have had a chance to remount that device, the chance never
>cropped up. So I went ahead with a test build - fully
>expecting it to fail when miniperl first invokes writemain.pl.
I think we knew this would work ok as long as ALWAYS_DEFTYPES is not defined, which it
now is not with a -des configure. There is something unusual about the stat/cando
combination in Perl_find_script in util.c that seems to kick in when ALWAYS_DEFTYPES
is defined; at least the cando() calls in there are where the problem cropped up
before, and they don't seem to even get invoked when ALWAYS_DEFTYPES is off.
There may well be other cases where the lib$fid_to_name code in vms.c gets hit
(potentially provoking the same issue) but I have not been able to come up with a
simple test case. Basically the following condition would have to evaluate to false
in Perl_cando for the lib$fid_to_name call to be invoked:
statbufp == &PL_statcache
If the passed stat buffer is always the same as the cache (which is usually the case),
then the lib$fid_to_name call will never happen, so it's difficult if not impossible
to reproduce the problem by just using file test macros. If anyone knows of a valid
circumstance where the above condition evaluates to false, I'd love to know about it.
On the other hand, if that circumstance should never happen, then why are we testing
for it?
____________________________________________
Craig A. Berry
mailto:[EMAIL PROTECTED]
