This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: PING [testsuite]: Tidy up testsuite handling of LD_LIBRARY_PATH
"H.J. Lu" <hjl.tools@gmail.com> writes:
> On Fri, Jul 10, 2009 at 11:36 AM, Richard
> Sandiford<rdsandiford@googlemail.com> wrote:
>> "H.J. Lu" <hjl.tools@gmail.com> writes:
>>> On Fri, Jul 10, 2009 at 2:52 AM, Richard
>>> Sandiford<rdsandiford@googlemail.com> wrote:
>>>> 2009/7/10 H.J. Lu <hjl.tools@gmail.com>:
>>>>> On Thu, Jul 9, 2009 at 7:49 PM, H.J. Lu<hjl.tools@gmail.com> wrote:
>>>>>> On Tue, Jun 30, 2009 at 3:20 PM, H.J. Lu<hjl.tools@gmail.com> wrote:
>>>>>>> On Tue, Jun 30, 2009 at 11:38 AM, Richard
>>>>>>> Sandiford<rdsandiford@googlemail.com> wrote:
>>>>>>>> Janis Johnson <janis187@us.ibm.com> writes:
>>>>>>>>> On Sat, 2009-06-27 at 09:42 +0100, Richard Sandiford wrote:
>>>>>>>>>> Ping for this patch:
>>>>>>>>>>
>>>>>>>>>> ? ? http://gcc.gnu.org/ml/gcc-patches/2009-06/msg00151.html
>>>>>>>>>>
>>>>>>>>>> (which should probably have had [testsuite] in the header, sorry).
>>>>>>>>>
>>>>>>>>> Good grief, it already had "testsuite" in the header twice! ?I don't
>>>>>>>>> know how I missed it.
>>>>>>>>>
>>>>>>>>> This is an area with which I'm not comfortable, but the patch looks
>>>>>>>>> much better thought-out than what was there before, so OK.
>>>>>>>>
>>>>>>>> Thanks, applied. ?I agree this is a bit of a tender area, so please
>>>>>>>> shout if I've broken something.
>>>>>>>>
>>>>>>>
>>>>>>> It may have caused:
>>>>>>>
>>>>>>> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40601
>>>>>>>
>>>>>>
>>>>>> It also caused:
>>>>>>
>>>>>> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40707
>>>>>>
>>>>>
>>>>> It also caused:
>>>>>
>>>>> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40709
>>>>
>>>> OK, it looks like the cure is worse than the disease. ?I'll revert
>>>> these patches when I get home.
>>>> Sorry once again for all the hassle.
>>>>
>>>
>>> I posted a patch for PR 40707:
>>>
>>> http://gcc.gnu.org/ml/gcc-patches/2009-07/msg00602.html
>>>
>>> and one for 40709:
>>>
>>> http://gcc.gnu.org/ml/gcc-patches/2009-07/msg00603.html
>>
>> Thanks HJ.
>>
>> TBH, I'm not sure I really follow what's going wrong in 40707.
>> How did you stop the pre-patch harness from trampling over
>> LD_LIBRARY_PATH? ?It could set the variable too, and it also
>> used "[is_remote target]".
>
> We set up ld_library_path, not LD_LIBRARY_PATH. It is up
> to board to set up LD_LIBRARY_PATH. See unix_load
> in /usr/share/dejagnu/config/unix.exp.
Sorry, I still don't follow. Before the patch, the harness used
ld_library_path to build up its own library path for the newly-built
libraries. And before the patch, set_ld_library_path_env_vars could
set the LD_LIBRARY_PATH environment variable to $ld_library_path.
Both should have remained true after the patch. The intent of the
patch was simply to remove directories for other multilibs from the
library path.
I'm not claiming I haven't broken something in 40707. I'm just not
sure specifically what it is that I broke. (And yeah, I'm probably
being dense here. ;) )
E.g., 40709 shows that we no longer set the ld_library_path in some
cases where we should have done. Was 40707 more fallout from that bug,
or was it something else?
The same change meant that we no longer called set_ld_library_path_env_vars
when c-torture.exp and gcc-dg.exp are sourced, but did it later on instead.
Was 40707 fallout from that instead?
Is your board description file something that you'd be prepared to post,
or is it confidential?
I realise it's probably academic to you given the end of this message,
but I'd still like to understand in case I do revisit this.
>> My instinct is still to revert the patch. ?It also caused PR40699,
>> although from Rainer's last message (thanks) it seems like it might
>> be tickling an underlying problem.
>
> I have no objections to revert the whole thing.
'Tis done!
Richard