Bug#289884: The Configure script seems buggy

[Marc Pignat]

On Saturday 22 May 2010 14:31:47 Niko Tyni wrote:
...
> Does the test program failure depend on /proc/cpu/alignment ?

Not really. Using the default value (0) for /proc/cpu/alignment, the program will fail with a bad
exit code (the read or write value is not good), but it seems that the configure script is
ignoring that value...

It seems to be a problem with the handling of the return value of the alignment
test, perhaps a shell problem?

> 
> I expect that fixing the problem will change the ABI, which is something
> we can't (easily) do for the 5.10.1 series. However, 5.12.1 is only in
> experimental and ABI changes are still possible for that branch.

Telling perl that it can't uses unaligned access on ARM machine will probably
make no ABI changes...

> 
> I wonder if the armel slowness reported in #543477 is related.
running "time perl -e 'use Getopt::Long;'" does not trigger alignment errors, so this seems unrelated!


> 
> I'm a bit short on time, will investigate properly later.
> 
I'm testing a patch, I will send it as soon asap. Unfortunately, it takes ages for
compiling on this target...

Regards


Marc