i agree, that in the long run we might need something better; we can
carry that forward upstream when pushing our patch there though. For now
in ubuntu the patch jamie proposed seems fine.
--
[armel] fio fails to build with ARMv7 optimizations
https://bugs.launchpad.net/bugs/532722
You received
Attaching .debdiff
** Attachment added: debdiff fix
http://launchpadlibrarian.net/40482727/fio_1.33.1-1ubuntu1.debdiff
--
[armel] fio fails to build with ARMv7 optimizations
https://bugs.launchpad.net/bugs/532722
You received this bug notification because you are a member of Ubuntu
Bugs,
OK, fair enough; I hadn't looked at the ifdefs preceding the patch.
Note that ARMv6 is missing: it may be a good idea to add some extra
checks:
#ifdef defined(__ARM_ARCH_4__) || defined(__ARM_ARCH_4T__) ||
defined(__ARM_ARCH_5__) defined(__ARM_ARCH_5T__) || defined(__ARM_ARCH_5TE__)
||
** Attachment added: repost with out the random trailing comment
http://launchpadlibrarian.net/40482956/cc_configure
--
[armel] fio fails to build with ARMv7 optimizations
https://bugs.launchpad.net/bugs/532722
You received this bug notification because you are a member of Ubuntu
Bugs, which
This bug was fixed in the package fio - 1.33.1-1ubuntu1
---
fio (1.33.1-1ubuntu1) lucid; urgency=low
* arch/arch-arm.h: Fix fail to build issue on ARMv7 and make it SMP
safe (LP: #532722).
-- Jamie Bennett ja...@linuxuk.org Thu, 04 Mar 2010 20:38:38 +
** Changed in: fio
Dave, i think adding that script to wiki would be helpful. Together with
an example macro that shows how to use it for folks not familiar with m4
etc.
If you could add that jamie would probably be happy to update his patch
to do it that way.
--
[armel] fio fails to build with ARMv7
Added more architecture defines as suggested by Dave Martin.
Fixed fail to build on ia64.
** Attachment added: fio_1.33.1-1ubuntu1.debdiff
http://launchpadlibrarian.net/40486269/fio_1.33.1-1ubuntu1.debdiff
--
[armel] fio fails to build with ARMv7 optimizations
Probably going to change this fix for the configure script Dave Martin
suggested, seems the more sensible approach.
--
[armel] fio fails to build with ARMv7 optimizations
https://bugs.launchpad.net/bugs/532722
You received this bug notification because you are a member of Ubuntu
Bugs, which is
My configure program Is just a shell script btw: comments at the start
of the script explain how to use it.
Hopefully this is a reasonable approach when a package hasn't moved to a
heavyweight autotools-based flow yet, but I'm open to suggestions...
--
[armel] fio fails to build with ARMv7
Some small suggestions on the shell script:
- mktemp /tmp/$me-XX = mktemp -t $me-
- instead of echo -n + final echo, why not just accumulate defines in a string
which you output? e.g. DEFINES=, DEFINES=$DEFINES -DFOO and echo CFLAGS
+= $DEFINES
- you could eval this from make
Some small suggestions on the shell script:
- mktemp /tmp/$me-XX = mktemp -t $me-
Hmmm, I learn something every day :) However, man mktemp says -t is
deprecated.
But mktemp -p /tmp $me- would work equally well.
- instead of echo -n + final echo, why not just accumulate
In the process of fixing this for ARMv7 I have added the correct calls
to ensure SMP safety.
** Patch added: fio-1.22.1.patch
http://launchpadlibrarian.net/40260566/fio-1.22.1.patch
** Tags added: armv7 patch
--
[armel] fio fails to build with ARMv7 optimizations
Tested on iMX51 ubuntu-netbook form 05/03/10
--
[armel] fio fails to build with ARMv7 optimizations
https://bugs.launchpad.net/bugs/532722
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
--
ubuntu-bugs mailing list
Of course that should of read Tested on the iMX51 ubuntu-netbook image
from 05/03/10
--
[armel] fio fails to build with ARMv7 optimizations
https://bugs.launchpad.net/bugs/532722
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
--
I have a concern about the proposed patch: if the code is ever built for
an arch 7-a (in the future) the wrong code will silently be built.
Is it not feasible to use configure tests to determine whether
asm(nop) and __sync_* are supported? This is more future-proof...
--
[armel] fio fails to
As long as __ARCH_ARM_7A__ isn't defined the build will fail as the
defines will not be declared.
New arch's will have to be added to the file in the future which is how
the packages does it now. We can investigate a better fix if desired.
Note that this package doesn't use configure so we are
16 matches
Mail list logo