Mailing List Archive

get_number_of_jobs
Why does get_number_of_jobs reduces the number of parallel jobs to "to
ensure successful merge"? In my humble opinion if a package fails to
compile with a large -j then the ebuild should know the limit and
reduce it.

An example of the problem: My machine has one processor but there are
many icecc daemons running on the local network. get_number_of_jobs
makes libstdc++-v3 compile with -j2 but It could use -j25 from the
/etc/make.conf. Removing the get_number_of_jobs makes libstdc++-v3
compile much faster.

Thanks for any comments,
Rafael Ávila de Espíndola

--
gentoo-dev@gentoo.org mailing list
Re: get_number_of_jobs [ In reply to ]
On Thursday 12 May 2005 02:40 pm, Rafael Espíndola wrote:
> Why does get_number_of_jobs

it's an old hack that should be punted ... it's only used by glibc/gcc now, i
just havent gotten around to removing it

there's some parallel build issues in glibc, i just havent gotten up the guts
to track it down (anyone who has peeked inside glibc code can understand
why :P)
-mike

--
gentoo-dev@gentoo.org mailing list
Re: get_number_of_jobs [ In reply to ]
Rafael Espíndola wrote: [Thu May 12 2005, 02:40:46PM EDT]
> Why does get_number_of_jobs reduces the number of parallel jobs to "to
> ensure successful merge"? In my humble opinion if a package fails to
> compile with a large -j then the ebuild should know the limit and
> reduce it.

Nope, if a package doesn't build with large -j then the ebuild should
force -j1. Either a package is parallel build clean or it isn't. You
can't make the problem go away by using an intermediate value for -j;
it might work for you but it won't for somebody else.

Regards,
Aron

--
Aron Griffis
Gentoo Linux Developer
Re: get_number_of_jobs [ In reply to ]
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Mike Frysinger wrote:
| On Thursday 12 May 2005 02:40 pm, Rafael Espíndola wrote:
|
|>Why does get_number_of_jobs
|
|
| it's an old hack that should be punted ... it's only used by glibc/gcc
now, i
| just havent gotten around to removing it

will be great when the limitation is gone, then i can stop stealing it's
code and putting it straight in an ebuild sans the evil job limiting
function_reuse++

|
| there's some parallel build issues in glibc, i just havent gotten up
the guts
| to track it down (anyone who has peeked inside glibc code can understand
| why :P)
| -mike
|

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.1 (GNU/Linux)
Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org

iD8DBQFChLMGUpKYMelfdYERAgkjAJ96WoVxxV0SK5epQxGpcR+QcqeBrgCfUvC/
XAr1J6tPTKAhe2OJ8TCwCZU=
=dp4e
-----END PGP SIGNATURE-----
--
gentoo-dev@gentoo.org mailing list