[Varnish] #344: varnish won't build on ppc/ppc64 using jemalloc

Varnish varnish-bugs at projects.linpro.no
Fri Oct 3 22:47:04 CEST 2008


#344: varnish won't build on ppc/ppc64 using jemalloc
--------------------+-------------------------------------------------------
 Reporter:  ingvar  |       Owner:       
     Type:  defect  |      Status:  new  
 Priority:  normal  |   Milestone:       
Component:  build   |     Version:  trunk
 Severity:  normal  |    Keywords:       
--------------------+-------------------------------------------------------
 As reported on IRC, there is something, uhm, strange going on when
 building varnish on the ppc64 boxes in koji, the build farm environment
 for Fedora. i386 and x86_64 builds work without problems. ppc and ppc64
 builds fails unless deactivating jemalloc.

 Then I dug out my old RS6000 B50, and put it on net in the attic, and
 installed "mock" on it. Mock is the same build environment as koji uses,
 or at least, that's what is't supposed to be. On the B50, varnish compiles
 with jemalloc in mock for fedora rawhide (soon-to-become fedora 10)
 without problems at all. (I could not test ppc64 as as that old box is
 only 32bit ppc.)

 Now, I whined a bit about this at #fedora-ppc, and got this, uhm,
 insightful answer from David Woodhouse, one of the glibc upstream
 maintainers:

 "It's possible that in koji we use 64KiB pages, rather than 4KiB. Is this
 the package with the 'special' allocator? Make it use getpagesize()
 instead of hard-coding the page size to some number it pulls out of its
 arse."

 Is it possible to do this?

 Earlier, Woodhouse has asked why we can't use the stock glibc malloc, and
 rather file a bug against glibc explaining what problems their malloc has.

 I would suggest to do this anyhow. A normal bug on Fedora in
 http://bugzilla.redhat.com/ would do, as the Fedora maintainers are also
 upstream maintainers. The worst thing that could happen would be that the
 reporter got flamed.

 Ingvar

-- 
Ticket URL: <http://varnish.projects.linpro.no/ticket/344>
Varnish <http://varnish.projects.linpro.no/>
The Varnish HTTP Accelerator


More information about the varnish-bugs mailing list