278307 - When GRE_HOME is set, #ifndef XP_WIN32 and #ifndef XP_UNIX we return null unexpectedly

Status: RESOLVED FIXED

(Core :: XPCOM, defect)

Description

[Justin Wood (:Callek)]

This coming just adds an #else case where we just hope for the best with the
setting of GRE_HOME, rather than returning a null static here.

Patch coming...

Comment 1

[Justin Wood (:Callek)]

Attachment: described fix

This is better than we have now...

Comment 2

[Darin Fisher]

Comment on attachment 171206 [details] [diff] [review]
described fix

ok, sure.  add a space between '//' and 'hope' and r+sr=darin

Comment 3

[Darin Fisher]

fixed-on-trunk

Comment 4

[Doug Turner (:dougt)]

Comment on attachment 171206 [details] [diff] [review]
described fix

what about the unix case?  do you care if realpath fails?  do you want to have
a 'hope for the best' case xp?

Comment 5

[Darin Fisher]

> what about the unix case?  do you care if realpath fails?  do you want to have
> a 'hope for the best' case xp?

that's a good question.  callek, any thoughts?

Comment 6

[Justin Wood (:Callek)]

This isnt really my area, and I only personally code/work on win32, but I would
assume if realpath fails then returning GRE_HOME would be a bad idea, since then
our code would assume the path is real when it was already checked and found to
be not.

if we want to set the static even if realpath fails, why bother with the check,
or even with the #if for that.  I would say the path check(s) are much better
than this #else cause, for any OS.  But the mere setting of GRE_HOME should not
have stripped any hope of making sGRELocation set.

...perhaps only the return if sGRELocation is set, and if not break through to
the following attempts, such as the win32 Registry check...

Any of these I would personally be happy with, as long as the simple setting of
GRE_HOME=anything does not make us fail sGRELocation on those OS's inertly as
before.