Hi Kurt,
Thanks for the Debian builds of elfutils-0.165. Sorry for the breakage, wrt old glibc elf.h system header. But happy to see the simple fix seems to work.
I looked at the latest build logs: https://buildd.debian.org/status/package.php?p=elfutils
For the hppa failure we need some hppa hacker to explain whether or not this is a real issue or not:
elflint /<<PKGBUILDDIR>>/tests/elfstrmerge section [24] '.plt' is both executable and writable FAIL run-strip-strmerge.sh (exit status: 1)
Normaly elflint is correct, this would be a bad thing.
The hurd test build issue should be easy to fix with the attached patch.
For the kFreeBSD variants you could start with the second patch to elflint to at least recognize ELFOSABI_FREEBSD. Although I suspect there might be more issues. But maybe we are lucky and it really is just like normal GNU/Linux ELF files since you are using the normal GNU toolchain.
The other issue on kFreeBSD is that we try to run a test that uses the linux specific dwfl_linux_proc_attach. We should just skip that on none GNU/Linux platforms like the third attached patch does.
The mips failures are a bit harder to fix since elflint really doesn't like some of the mips specific issues. We really should import the Debian mips backend, I am sorry I haven't found time to do all the needed research.
If you could try the attached patches on a new Debian build that would be appreciated.
Thanks,
Mark
On Wed, Jan 13, 2016 at 11:13:28PM +0100, Mark Wielaard wrote:
Hi Kurt,
Thanks for the Debian builds of elfutils-0.165. Sorry for the breakage, wrt old glibc elf.h system header. But happy to see the simple fix seems to work.
I looked at the latest build logs: https://buildd.debian.org/status/package.php?p=elfutils
For the hppa failure we need some hppa hacker to explain whether or not this is a real issue or not:
elflint /<<PKGBUILDDIR>>/tests/elfstrmerge section [24] '.plt' is both executable and writable FAIL run-strip-strmerge.sh (exit status: 1)
Normaly elflint is correct, this would be a bad thing.
I think that has come up before in the past.
For the kFreeBSD variants you could start with the second patch to elflint to at least recognize ELFOSABI_FREEBSD. Although I suspect there might be more issues. But maybe we are lucky and it really is just like normal GNU/Linux ELF files since you are using the normal GNU toolchain.
It has always worked without problems.
If you could try the attached patches on a new Debian build that would be appreciated.
I'll try it tomorrow or in the weekend.
Kurt
On Thu, 2016-01-14 at 00:22 +0100, Kurt Roeckx wrote:
On Wed, Jan 13, 2016 at 11:13:28PM +0100, Mark Wielaard wrote:
I looked at the latest build logs: https://buildd.debian.org/status/package.php?p=elfutils [...] If you could try the attached patches on a new Debian build that would be appreciated.
I'll try it tomorrow or in the weekend.
Thanks for doing a build with the proposed patches. They seem to have solved the issues on kfreebsd and the hurd port builds again (but still fails one testcase). I have pushed the commits below to master.
The only thing that worries me is the sparc failure. I have CCed Jose who might want to take a look. What is strange is that 0.165-1 and 0.165-2 PASS on sparc just fine. And I don't see anything in 0.165-3 that would explain the failure on sparc: https://buildd.debian.org/status/logs.php?pkg=elfutils&arch=sparc64
Thanks,
Mark
commit 0cd02dcafcfdaff56f483f41c0ec45de756c7083 Author: Mark Wielaard mjw@redhat.com Date: Wed Jan 13 23:06:33 2016 +0100
tests: Skip dwfl-bug-fd-leak test if dwfl_linux_proc_report is unsupported.
Signed-off-by: Mark Wielaard mjw@redhat.com
commit 624e3cb81d436282fa9f6ce4fa63203bad236057 Author: Mark Wielaard mjw@redhat.com Date: Wed Jan 13 22:57:03 2016 +0100
elflint: Recognize ELFOSABI_FREEBSD which Debian kFreeBSD uses.
Signed-off-by: Mark Wielaard mjw@redhat.com
commit d460773dd4fdeb98d9271f741e0a54f05ced876d Author: Mark Wielaard mjw@redhat.com Date: Wed Jan 13 22:49:02 2016 +0100
tests: Guard linux specific header includes with ifdef __linux__.
Signed-off-by: Mark Wielaard mjw@redhat.com
On Thu, 2016-01-14 at 00:22 +0100, Kurt Roeckx wrote: > On Wed, Jan 13, 2016 at 11:13:28PM +0100, Mark Wielaard wrote: > > I looked at the latest build logs: > > https://buildd.debian.org/status/package.php?p=elfutils > > [...] > > If you could try the attached patches on a new Debian build that would > > be appreciated. > > I'll try it tomorrow or in the weekend.
Thanks for doing a build with the proposed patches. They seem to have solved the issues on kfreebsd and the hurd port builds again (but still fails one testcase). I have pushed the commits below to master.
The only thing that worries me is the sparc failure. I have CCed Jose who might want to take a look. What is strange is that 0.165-1 and 0.165-2 PASS on sparc just fine. And I don't see anything in 0.165-3 that would explain the failure on sparc: https://buildd.debian.org/status/logs.php?pkg=elfutils&arch=sparc64
The problem seems to be:
/«PKGBUILDDIR»/tests/backtrace: dwfl_thread_getframes: No DWARF information found
and that is probably triggered by this code in __libdwfl_frame_unwind:
Dwfl_Module *mod = INTUSE(dwfl_addrmodule) (state->thread->process->dwfl, pc); if (mod == NULL) __libdwfl_seterrno (DWFL_E_NO_DWARF);
I definitely can't reproduce this problem in my own sparc64-*-* setups.
On Mon, Jan 18, 2016 at 03:35:47PM +0100, Mark Wielaard wrote:
On Thu, 2016-01-14 at 00:22 +0100, Kurt Roeckx wrote:
On Wed, Jan 13, 2016 at 11:13:28PM +0100, Mark Wielaard wrote:
I looked at the latest build logs: https://buildd.debian.org/status/package.php?p=elfutils [...] If you could try the attached patches on a new Debian build that would be appreciated.
I'll try it tomorrow or in the weekend.
Thanks for doing a build with the proposed patches. They seem to have solved the issues on kfreebsd and the hurd port builds again (but still fails one testcase). I have pushed the commits below to master.
The only thing that worries me is the sparc failure. I have CCed Jose who might want to take a look. What is strange is that 0.165-1 and 0.165-2 PASS on sparc just fine. And I don't see anything in 0.165-3 that would explain the failure on sparc: https://buildd.debian.org/status/logs.php?pkg=elfutils&arch=sparc64
So I let it try again on sparc and it worked now. No idea what is going on.
Kurt
On Wed, Jan 13, 2016 at 11:13:28PM +0100, Mark Wielaard wrote:
The mips failures are a bit harder to fix since elflint really doesn't like some of the mips specific issues. We really should import the Debian mips backend, I am sorry I haven't found time to do all the needed research.
So at least one of the mips problems is that the current patch just does: RELOC_TYPE (NONE, 0) RELOC_TYPE (16, 0) RELOC_TYPE (32, 0) RELOC_TYPE (REL32, 0) RELOC_TYPE (26, 0) RELOC_TYPE (HI16, 0) RELOC_TYPE (LO16, 0) [...]
Kurt
elfutils-devel@lists.fedorahosted.org