Re: Regressions when building glibc for HPPA.
On Fri, May 21, 2010 at 5:06 PM, Carlos O'Donell
<carlos@systemhalted.org> wrote:
> On Tue, May 18, 2010 at 5:32 PM, Aurelien Jarno <aurel32@debian.org> wrote:
>> Any progress on that? We are approaching the transition freeze deadline,
>> so we will need to upload eglibc 2.11 to unstable really soon now.
>
> Sorry, I forgot to send out an update. I made some progress
> understanding the one of the failures, but I don't have a fix yet. I
> plan to spend Saturday and Sunday debugging the failures. I will keep
> you updated.
There are 5 regressions for the eglibc 2.11-0exp7 release.
(1) tstdiomisc
(2) tst-fork1
(3) tststatic
(4) tststatic2
(5) tst-tls9-static
Test (1) is a compiler bug involving the multiplication of -1 * NAN.
The compiler defect causes the test to fail. The actual stdio
implementation in glibc is working correctly. I have manually tested
the stdio functions and they work correctly. This test can be XFAIL'd.
Test (2) is a VIPT-WB cache issue in the kernel. The HPPA port has a
patch set that is currently being worked on to fix this cache problem,
until then I expect tst-fork1 and tst-fork2 to fail intermittently.
There for tst-fork1 and tst-fork2 should be XFAIL'd until the kernel
bug is fixed. There is nothing we can do until then.
I am still analyzing (3), (4) and (5) which appears to be related.
Cheers,
Carlos.
Reply to: