Package: libtool
Version: 2.4.6-2
Severity: important
Tags: upstream fixed-upstream
Forwarded: https://debbugs.gnu.org/cgi/bugreport.cgi?bug=20006

Dear Maintainer,

libtool 2.4.6-0.1 has introduced a performance regression which results
in significant slowdown in compile time. Version 2.4.6-2 is still
affected.

Especially buildds of slower arches like mips* and arm* suffer a lot
from this performance regression. Here are a few examples how mesa is
affected by this bug:

[2016-02-07] Accepted libtool 2.4.6-0.1 (source amd64 all) into unstable

mipsel: https://buildd.debian.org/status/logs.php?pkg=mesa&arch=mipsel
11.1.1-2 (sid)      Maybe-Successful    2016-01-15 19:04:32     mipsel-aql-02   
50m
11.1.2-1 (sid)      Maybe-Successful    2016-02-11 14:30:46     mipsel-aql-02   
1h 22m

armhf: https://buildd.debian.org/status/logs.php?pkg=mesa&arch=armhf
11.1.1-2 (sid)      Maybe-Successful    2016-01-15 19:15:32     hoiby   34m
11.1.2-1 (sid)      Maybe-Successful    2016-02-11 12:39:21     hoiby   52m

i386: https://buildd.debian.org/status/logs.php?pkg=mesa&arch=i386
11.1.1-2 (sid)      Maybe-Successful    2016-01-15 18:20:15     x86-grnet-01    
7m
11.1.2-1 (sid)      Maybe-Successful    2016-02-11 11:54:33     x86-grnet-01    
10m

x32: https://buildd.debian.org/status/logs.php?pkg=mesa&arch=x32
11.1.1-2 (sid)      Maybe-Successful    2016-01-15 18:36:30     physik  17m
11.1.2-1 (sid)      Maybe-Successful    2016-02-11 12:36:45     physik  28m

or systemd on x32: 
https://buildd.debian.org/status/logs.php?pkg=systemd&arch=x32
228-6 (sid)         Maybe-Successful    2016-02-04 12:42:54     physik  42m
229-1 (sid)         Maybe-Successful    2016-02-11 23:35:48     physik  1h 3m


This bug has been reported to upstream and has been fixed already [1].
There's also a blog post about this issue [2].
Unfortunately upstream hasn't released a new upstream release with those
fixes yet.

Please cherry-pick the following three patches from upstream master
branch to mitigate the excessive sed forking:

http://git.savannah.gnu.org/cgit/libtool.git/commit/?id=32f0df9835ac15ac17e04be57c368172c3ad1d19
http://git.savannah.gnu.org/cgit/libtool.git/commit/?id=16dbc070d32e6d4601cb5878dfdf69f2e29c84e1
http://git.savannah.gnu.org/cgit/libtool.git/commit/?id=9187e9a231e0a06cc29c336857e95f07f855b2c9

With those patches applied, libtool 2.4.6 still isn't as fast as 2.4.2
was, but it's much better than the unpatched version.


Thanks,
Andreas

[1] https://debbugs.gnu.org/cgi/bugreport.cgi?bug=20006
[2] https://harald.hoyer.xyz/2015/03/05/libtool-getting-rid-of-180000-sed-forks/

Attachment: signature.asc
Description: Digital signature

Reply via email to