Please test: patch that makes the build script work in parallel, improves build time

Kévin Raymond shaiton at fedoraproject.org
Fri Apr 29 09:04:16 UTC 2011


On Fri, Apr 29, 2011 at 10:03 AM, Elad <el.il at doom.co.il> wrote:
> These patches[1][2] makes it possible to run our build script with the
> -j parameter, which makes it use more then one process and thus
> improves build time on multi-core systems.
> to test, run make -j2 make -j3 or make -j4 (the number is the number
> of allowed jobs, shouldn't be higher then your number of cores). don't
> run it with -j without a number, because it will run as many jobs as
> it can, resulting your machine to become un-responsive until the build
> is complete (that's was the result on my machine, might differ on
> others).
> Also it won't affect if you are building single language, it only
> improves speed if you build two or more.
>
> The second patch fixes a bug that Sijis discovered yesterday with the
> first patch, related to RSS caching.
> Don't forget to run make clean before you test.
>
> I won't push it before it gets tested properly. One of the reason it
> requires a lot of testing is that I don't know python and this is the
> first time I'm editing Makefiles, so I probably did some mistakes.
>
> During my attempts to fix that bug, I discovered that there is a
> forgotten test file, fedoraproject.org/data/content/rss.html that
> pulls the redhat magazine feed, which takes long to download (longer
> than the fedora feeds). As we don't really use this feed, I suggest
> deleting this file for more speed improvements.
>
> [1] http://elad.fedorapeople.org/0001-Buildscript-can-now-work-parallel-improves-build-tim.patch
> [2] http://elad.fedorapeople.org/0002-Fix-RSS-caching-bug-in-parallel.patch
> --
> -Elad.
> --
> websites mailing list
> websites at lists.fedoraproject.org
> https://admin.fedoraproject.org/mailman/listinfo/websites
>

Hi guys,

Thanks Elad for improving the build time.
Applying both patches works great for me (in the master branch...):

make clean && make all
real	3m6.863s

make clean && make all
real 0m54.836s
(AMD PhenomII X4 955)

+1 to apply

+1 to remove the data/content/rss.html RH feed parsing.

msgfmt in concurrency also would be nice (just to know that our build
process is perfect:) )


-- 
Kévin Raymond
User:shaiton
GPG-Key: A5BCB3A2



More information about the websites mailing list