[ft] Optimizer problem? (FT2.3.7 + GCC4.2.3)

[Ian Britten]

Hi all,
I'm just fishing for feedback at this time, about whether anyone is
using FT v2.3.7 with GCC v4.2.3, and whether they're seeing any
problems?

<Background>
I have two virtually identical Linux machines:
 1) Mandriva 2008.1, 32-bit
 2) Mandriva 2008.1, 64-bit

When I run our unit tests on box #2, there are no problems, and
Valgrind reports no problems.  As well, there's no problems running
our tests on Windows, which reinforces my thoughts that our code
is correct.

When I run our unit tests on box #1, they crash, down inside
FTC_CMapCache_Lookup()  (Details below)
</Background>


<Details>
In the case that crashes, it's in FTC_CMapCache_Lookup(), trying
to lookup char_code 79 of the standard Bitstream Vera TT font.

It *seems* that FTC_Manager_LookupFace() returned either an
uninitialized or dangling FT_Face pointer, which when used
further down in that function, results in a crash.

When it crashes, the only problem reported by Valgrind is the
use of the uninitialized pointer, within FTC_CMapCache_Lookup()
</Details>


On these boxes, I'm using the system-installed version of FT2.3.7,
which is obviously built optimized.  If I build a debug version of
FT, and relink my unit-test program against it, all problems go
away - No crash, no Valgrind errors.

I'm torn as to whether it's possibly a problem with our code (eg:
How it's using the FT cache) that only appears on one platform, or
whether it's maybe another problem similar to
https://savannah.nongnu.org/bugs/index.php?23556
Given that our unit tests work problem-free on a couple of other
platforms (64-bit Linux and Windows), I'm beginning to suspect it
may be the latter.

I apologize that I don't have an isolated reproducible FT test
case at this time.  (And, it'll take some work to isolate just
the involved FT calls out of our tests).
Anyways, if anyone has any thoughts/suggestions/etc,it'd be
appreciated!
Ian