Minimum supported version of Forte

[Christopher Seawood]

Unless anyone has any major objections, we're going to start treating
the Forte 6U2 release as the minimal supported version of the Forte
compiler that we can use to compile Mozilla. Forte 6U2 is what the
nebiros tinderbox is using. I don't plan on adding any checks to
enforce this change unless it becomes a problem. We're also going to
bump the default optimization level for Forte builds to -xO4.

http://bugzilla.mozilla.org/show_bug.cgi?id=109860

- cls

[Matthew Langford]

Post by Christopher Seawood
Unless anyone has any major objections, we're going to start treating
the Forte 6U2 release as the minimal supported version of the Forte
compiler that we can use to compile Mozilla. Forte 6U2 is what the
nebiros tinderbox is using. I don't plan on adding any checks to
enforce this change unless it becomes a problem. We're also going to
bump the default optimization level for Forte builds to -xO4.
http://bugzilla.mozilla.org/show_bug.cgi?id=109860

Not to be too pessimistic, but if I compile glib 2.0.6, using Forte 6U2,
with anything higher than -xO2, I get a segfault in the glib date routine.
If higher than -xO3, segfaults in both date test and in string (printf)
tests. Yech. This with up-to-date patching.

Is mozilla less likely to run into these potential compiler bugs? Or am I
tickling bugs no one else sees? Of course, all of my Forte builds of
mozilla have only segfaulted to date.

--
MattLangford

[Christopher Seawood]

Post by Matthew Langford

Post by Christopher Seawood
bump the default optimization level for Forte builds to -xO4.
http://bugzilla.mozilla.org/show_bug.cgi?id=109860

Not to be too pessimistic, but if I compile glib 2.0.6, using Forte 6U2,
with anything higher than -xO2, I get a segfault in the glib date routine.
If higher than -xO3, segfaults in both date test and in string (printf)
tests. Yech. This with up-to-date patching.
Is mozilla less likely to run into these potential compiler bugs? Or am I
tickling bugs no one else sees? Of course, all of my Forte builds of
mozilla have only segfaulted to date.

Without any specifics about the crashes or the code used by glib 2.x, I
couldn't say. We were running a tinderbox using WS6U2 + system updates
for about 3 months and the runtime tests worked fine. The Sun engineers
claim that -xO5 with 6U1 is what they use on their build machines so if
there were any problems, I would have expected them to be caught by now.

- cls