Just tested building on armhf with a 2.6.38 kernel here. The bug is
"fixed", as it it does not generate SIGILL. So the package is buildable
on armhf, but I do agree with Dave here, the method to detect float-
alignment should probably be changed.
Konstantinos
--
You received this bug notificatio
Maybe this bug can be closed, but in that case some bug should be kept
open for tracking purposes -- the build-time test done by H5detect still
appears wrong by design, and this probably needs to be addressed at some
point.
--
You received this bug notification because you are a member of Ubuntu
Yes it probably has to do with the kernel now. Sure the bug can be close,
but I'll be posting
any progress with it on armhf here anyway.
Konstantinos
On 18 February 2011 12:46, Jani Monoses wrote:
> Konstantinos I use the current omap4/natty kernel .35 and it worked, I
> don't know what version
Konstantinos I use the current omap4/natty kernel .35 and it worked, I
don't know what version runs on the buildd though.
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/635199
Title:
[armel] the ./H5
Just tried this on armhf, but still it fails with the same error
(admittedly the kernel there is 2.6.31, which explains it), I will try
again using a .38-pre kernel once I get my hands on a working one.
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscr
(argh, hit "post" by accident)
This probably won't generate a fault, because only integer accesses are
done to the target pointer in the generated code:
main () {
printf("%f\n", *(double *)((char *)main + 2));
}
This probably will (but it depends on compiler options, and on what code
the com
There's also the problem that deferencing a float * or double * pointer
won't necessarily generate an alignment fault.
This won't:
main () {
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/635199
Titl
What is the kernel on the buildds?
If it is >= 2.6.36, the fix to generate SIGBUS for alignment faults will
be present in the kernel. If the buildds are all >= 2.6.36, this may
solve the problem. If some are older, we might randomly get build
problems when the package gets rebuilt.
However, as
** Changed in: gcc-4.4 (Ubuntu)
Status: Confirmed => Invalid
** Changed in: hdf5 (Ubuntu)
Status: Confirmed => Fix Released
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/635199
Title:
I have just uploaded a new revision today restoring the build to armv7 and it
worked. The previous failure may have been specific to the maverick toolchain?
Is there a reason not to close this bug?
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed
I still think it would be reasonable to hard-code the results for float
alignment requirements on ARM.
On all ARM floating-point architecture versions, float and double must
be size-aligned. I don't think there has ever been any exception to
this rule, nor is any exception likely in the forseeabl
** Tags added: arm-porting-queue
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/635199
Title:
[armel] the ./H5detect test segfaults when built for armv7
--
ubuntu-bugs mailing list
ubuntu-bugs@list
I am not sure how this bug can be solved, Paul's comments seem to indicate that
there's no good test for this.
If it is a FTBFS with hardfloat (which is not yet supported in Ubuntu)
Konstantinos can you offer a solution?
The app may rely on the behaviour it tests for so I am not sure whether maki
Not on hardfloat. It still breaks regardless of arch.
On 19 January 2011 22:21, Jani Monoses wrote:
> If this is not also a toolchain issue then the gcc-4.4 bugtask can be closed.
> The goal is here to have the package built for armv7 and using armv5 for the
> specific test only?
> The original
If this is not also a toolchain issue then the gcc-4.4 bugtask can be closed.
The goal is here to have the package built for armv7 and using armv5 for the
specific test only?
The original issue was prompted by a FTBFS which was solved by building for
armv5.
--
You received this bug notification
** Changed in: hdf5 (Ubuntu)
Assignee: David Sugar (dyfet) => Jani Monoses (jani)
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/635199
Title:
[armel] the ./H5detect test segfaults when built fo
It's actually worse than that. The test can give incorrect results even for
the current machine. e.g.:
void foo(float * p) // assume p is misaligned
{
*p = 0; // first store
frob(0);
*p = 0; // second store
}
It is possible for the first store to succeed and the second to cause an
alignment faul
You are correct in that the test is done at "runtime", but of course
this actually only happens once at "configure" time for the package,
which means it is using the runtime characteristics of a specific
archive build machine, and at the time only. The behavior just happens
to currently appear rel
I believe the SIGILL is actually a red herring.
The underlying issue that the package appears to be testing for
misaligned access support at runtime. On ARM you can't do this
reliably. Behavior of misaligned accesses depends on both system
configuration and instruction choice. i.e. it may veys
On 22.09.2010 18:07, Dave Martin wrote:
> Shouldn't be an issue for maverick though, and shouldn't matter at all
> if the kernel is eventually fixed to generate SIGBUS.
could you open and track a kernel task?
--
[armel] the ./H5detect test segfaults when built for armv7
https://bugs.launchpad.ne
Hard-coding the alignment detection result for certain types on arm also
works --- this allows hdf5 to be built for armv7+vfpv3, and seems
reasonable since the alignment requirements for those types are common
across all arm architecture versions.
Note that if we care about this package working on
It is not a toolchain issue here. In H5detect.c:ALIGNMENT, a testing to
mis-aligned access is performed, as below,
*((TYPE*)(_buf+align_g[_ano])) = _val; /*possible SIGBUS or SEGSEGV*/
\
_val2 = *((TYPE*)(_buf+align_g[_ano])); /*possible SIGBUS or SEGSEGV*/
\
and SIGBUS and
** Changed in: gcc-4.4 (Ubuntu)
Assignee: (unassigned) => Yao Qi (yao-codesourcery)
--
[armel] the ./H5detect test segfaults when built for armv7
https://bugs.launchpad.net/bugs/635199
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
** Tags added: arm-mode
--
[armel] the ./H5detect test segfaults when built for armv7
https://bugs.launchpad.net/bugs/635199
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
--
ubuntu-bugs mailing list
ubuntu-bugs@lists.ubuntu.com
https:
thanks! reopening, and adjusting the milestone
** Changed in: hdf5 (Ubuntu)
Status: Fix Released => Confirmed
** Changed in: hdf5 (Ubuntu)
Milestone: ubuntu-10.10 => maverick-updates
--
[armel] the ./H5detect test segfaults when built for armv7
https://bugs.launchpad.net/bugs/635199
I think we only want to do this for building h5detect itself, and not to
use armv5t for the package as a whole. In fact, I was doing a small
patch to do something similar for h5detect alone.
--
[armel] the ./H5detect test segfaults when built for armv7
https://bugs.launchpad.net/bugs/635199
You
hdf5 (1.8.4-patch1-2ubuntu1) maverick; urgency=low
* On armel, build with -march=armv5t to avoid illegal instruction
in H5detect test. LP: #635199.
-- Matthias Klose Fri, 10 Sep 2010 18:46:05 +0200
** Changed in: hdf5 (Ubuntu)
Status: Confirmed => Fix Released
--
[armel] the .
david is tasked with fixing hdf5 in the ubuntu-arm team currently,
assigning to him
** Changed in: hdf5 (Ubuntu)
Assignee: (unassigned) => David Sugar (dyfet)
** Changed in: hdf5 (Ubuntu)
Milestone: None => ubuntu-10.10
--
[armel] the ./H5detect test segfaults when built for armv7
http
** Branch linked: lp:ubuntu/hdf5
--
[armel] the ./H5detect test segfaults when built for armv7
https://bugs.launchpad.net/bugs/635199
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
--
ubuntu-bugs mailing list
ubuntu-bugs@lists.ubuntu.c
see with 4.4.4-14ubuntu1
** Changed in: gcc-4.4 (Ubuntu)
Importance: Undecided => High
** Changed in: gcc-4.4 (Ubuntu)
Status: New => Confirmed
--
[armel] the ./H5detect test segfaults when built for armv7
https://bugs.launchpad.net/bugs/635199
You received this bug notification beca
30 matches
Mail list logo
