On (2013年04月02日 01:06), Mike Hommey wrote:
> The
> worst part (which is unfiled at the moment) is that PGO builds OOM on
> x86. That might require switching to cross-builds for x86. Or if we are
> lucky, only a more recent binutils.
> 
> 4.8 is IMHO too young to even consider testing.
> 
> Mike

Any hint to resolve OOM issue for x86 build?

I have seen this on my linux PC, and cannot proceed any more :-(

I posted about this in a posting to mozilla.dev.builds:
"(non-fatal?) Memory exhaustion of linux 32 build of comm-central debug
build during libxul.so link?"
on Mar 29.

It was non fatal back then.
Now, after a refresh of comm-central, just to see what goes on, it no longer
builds. Before, I could retry and somehow "ld" part proceeded
the second time. Not any more :-(

I checked my ulimit setting, swap size (2.5G external + main memory 3.6 G
allocated in VMPlayer), and ld version (timestamped in 2011.)
[I just noticed that in a follow-up post that LTO pass requires 6GB of
memory on 64bits linux, and so maybe I may be able to go on with a few more
memory in 64 bits environment, but there was a reason to stick to 32 bit
environment for me for non-mozilla related reasons.]

I suspect that that it may indeed have something to do with "ld" not
handling some situations gracefully.

I am using Debian GNU/Linux 32 bit distribution, and its ld
is
$ ld --version
GNU ld (GNU Binutils for Debian) 2.22


$ gcc --version
gcc (Debian 4.7.2-5) 4.7.2


BTW, you mentioned that GCC was switched on trybuilder.

It is not yet switched on Thunderbird TryServer, is it?
(I am not sure if I have seen this problem on Thunderbird last time I
checked. A few days ago...)

TIA

PS: I will post my finding if my switching to a newer binutils
would help. (2.23.1 seems to be the version. Its News does not suggest
anything remotely related to this issue.)
Or downgrading to gcc 4.6 might help.



_______________________________________________
dev-platform mailing list
dev-platform@lists.mozilla.org
https://lists.mozilla.org/listinfo/dev-platform

Reply via email to