[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: netscape?

> Oh, I believe it. Red-Hat compiles of mozilla run fine though.
> After installing, run once as root, and it will run for user accounts.

RH 6.x still used egcs 1.1.2, IIRC.  While this version of the compiler is
still broken WRT multiple-inheritence C++, it's not nearly as bad as gcc
2.95.x has been.  RH 7.x uses gcc "2.96", which has most of the patches
included that are needed to get C++ code compiling correctly.

> I've also been trying to compile mozilla with g++-3.0 and -O3 and
> 5 DAYS later it is still compiling.

Hehehe...only takes an hour or two on my UP2k at work :-P  I'll have to try
again with the more recent CVS snapshots (haven't done so yet).  I'm hoping
to get a good package uploaded sooner or later.

> Numerous problems exist, like mangled
> function names are different (fixed that by editing the genstubs.pl) and
> problems linking with the accessibility functions (fixed that by just
> making them return an error).

Can you send me a patch for the mangling fixes?  I was going to do this
exact procedure, but just haven't gotten to it yet (been busy porting
drivers and things to 2.4 kernels).  I wouldn't mind getting a peek at what
you did with the accessibility stuff as well, although I'll probably fix it
rather than just returning an error (if possible).

> On installing the deb package, registering components gives a
> segfault on my box. It's been doing that for months, and I strongly
> believe it's the compiler's fault.

It definitely is.  I went through several hours of debugging that and
arrived at the same conclusion pointed out by a fellow list-member: it's
definitely the compiler.  Seems that the CTORS and DTORS lists weren't being
set up properly, so when the exit functions were run and the binary tried to
call the destructors, it jumped immediately to the gp for the programme
itself (rather than iterating through the respective destructors).  For
simple C++ code, things work ok, but anything with multiple inheritence, it
obviously breaks.

> Lots of luck...



Reply to: