[gccsdk] Dynamic libraries vs pthreads

Peter Naulls peter at chocky.org
Tue Sep 9 19:28:18 PDT 2008

Lee Noar wrote:
> Peter Naulls wrote:
>> Lee Noar wrote:
>>> I've just bought an Iyonix, so when I've got it set up properly, I'll 
>>> have a closer look at this and the shared library version of Firefox.
> Just committed r3538 which fixes the QUIT signal in the program you 
> posted on 03/09/08. Turned out to be an "Alarm clock" signal which was 
> failing because it expected the address of the handler to be within 
> application space. Unfortunately, it probably won't fix anything in 
> Firefox because alarms are disabled in the desktop anyway.

Ok, excellent in any case.  From the same test, I have found
that the pthread_wait centisecond timer was instead using milliseconds
as translated from the nanosecond component, as converted from
microseconds to pass it (can no one agree on units ;-).  The upshot
was that a timeout could take up to 10 seconds longer than expected.

Will check in later and check for similar errors.

>> Here's something that everyone can try more easily - nspr and its
>> tests:
>>  cvs -d :pserver:anonymous at cvs-mirror.mozilla.org:/cvsroot co 
>> -rFIREFOX_2_0_0_17_RELEASE mozilla/nsprpub
>>  cd mozilla/nsprpub
>> Apply the patch from AB (make sure your SVN is up to date), e.g.:
>> patch -p1 < /usr/src/gccsdk/autobuilder/browser/firefox2/nspr-diff.patch
>> Now build:
>>  /home/riscos/env/ro-config
>>  /home/riscos/env/ro-make
> I'm having trouble building this due to a missing ro-ar command. I've 
> updated from svn and reapplied install-env - is this a script I should 
> have in /home/riscos/env ?

I meant the Autobuilder SVN up to date - ro-ar was a tool from the GCC 3
porting  tools which is no longer relevant, and hasn't been used for the
the Firefox builds for some considerable time.

More information about the gcc mailing list