I should be a bit more specific about why I don’t think it is a 
real “fix”. 
It works for the new pie enforcing debian version of gcc. We can put
that as a fact. It does so by replacing “-Wl,-r” by “-r”, so instead
of passing “-r” to the linker, gcc is now using a “-r” option directly
that passes the right magic to the linker presumably.

Presumably because I don’t have any documentation about “-r” in the version
of gcc 5.4.0 I have installed here. 

So, is this a new gcc 6+ option, an undocumented option or a specific
option of that debian compiler? In 2 out of 3 cases that means adopting
the debian fix will break on other compiler/distro.

Which may be why they didn’t follow it up to upstream flint/arb. It is a
debian specific patch. If this is the case the fix can only be applied
conditionally.

I may be poking myself in the eye here but I think just adopting the patch
is risky.

François

> On 6/11/2016, at 23:29, Francois Bissey <[email protected]> 
> wrote:
> 
> I don’t like the debian fix. In fact it doesn’t look like a fix at all.
> I would need a serious explanation of their change.
> That “-Wl,-r” already caused trouble in spack last month (on OS X) and as it 
> turns out it was spack's compiler wrappers’ fault. I am inclined to say
> that debian did something nasty or not well thought about to gcc.
> 
> Which feels rather improbable on the part of debian but we never know.
> 
> François
> 
>> On 6/11/2016, at 23:15, Emmanuel Charpentier <[email protected]> 
>> wrote:
>> 
>> After manually upgrading pip, I still get a raft of failed test. It seems 
>> (see uploaded log) that those are due to pip advertising a deprecation of 
>> teh format used for the database.
>> 
>> One errot (on simplicial_complex.py) is genuine ; this test appasses with no 
>> errors when run standalone (a long standing problem with this test.
>> 
>> I conclude that :
>>    - the patches pointed by Dima do solve the problem of compiling flint and 
>> arb on Debian testing
>>    - another problem is introduced by pip advertising.
>> 
>> Questions :
>>    - Should I package the patches that solve Trac#21782
>>    - Should I create a new ticket about  pip's newfound verbosity ?
>> 
>> HTH,
>> 
>> --
>> Emmanuel Charpentier
>> 
>> Le dimanche 6 novembre 2016 09:40:15 UTC+1, Emmanuel Charpentier a écrit :
>> A second attempt at rebuilding after distcleant gives the same results : a 
>> raft of errors mostly bound to pip's advertising. However, I an not sure of 
>> what I read in the results of tolerance.py...
>> 
>> I'll re-upgrade pip and post the results.
>> 
>> HTH,
>> 
>> --
>> Emmanuel Charpentier
>> 
>> 
>> Le dimanche 6 novembre 2016 03:20:26 UTC+1, Emmanuel Charpentier a écrit :
>> Sorry for being late : I had a severe NMI from the Real World (TM).
>> 
>> Using the patches pointed by Dima succeed in compiling Sage. However, I dat 
>> a raft of failutes which mostly relate to pip being outdated.
>> 
>> Updating pip "by hand" (./sage -pip install --upgrade pip) was a bad mistake 
>> : different errors at the same point,
>> 
>> I'll (make distclean && make ptestlong ) again  overnight and post the 
>> hopefully clean) log tomorrow.
>> 
>> HTH,
>> 
>> PS : about penalties for abusing compilers : did you consider this ?
>> 
>> --
>> Emmanuel Charpentier
>> 
>> Le dimanche 6 novembre 2016 01:04:36 UTC+1, François a écrit :
>> Apologies if you missed my later post. I had read the patch in reverse. 
>> You do the correct thing and they cripple it. In a curious way too. 
>> 
>> I suspect they do PIE (position independent executable - a relative of PIC) 
>> wrong. 
>> 
>> +1 about having severe penalties for crippling compilers. 
>> 
>> François 
>> 
>>> On 6/11/2016, at 12:19, 'Bill Hart' via sage-release 
>>> <[email protected]> wrote: 
>>> 
>>> On 5 November 2016 at 20:11, Francois Bissey <[email protected]> 
>>> wrote: 
>>> That was clearly a mistake in flint/arb in the first place. “-r” is a flag 
>>> for the linker, not sure what gcc was doing with it in the first place. 
>>> 
>>> -Wl is supposed to pass the option to the linker.   
>>> 
>> 
>> 
>> -- 
>> You received this message because you are subscribed to the Google Groups 
>> "sage-release" group.
>> To unsubscribe from this group and stop receiving emails from it, send an 
>> email to [email protected].
>> To post to this group, send email to [email protected].
>> Visit this group at https://groups.google.com/group/sage-release.
>> For more options, visit https://groups.google.com/d/optout.
>> <ptestlong.log>
> 
> -- 
> You received this message because you are subscribed to the Google Groups 
> "sage-release" group.
> To unsubscribe from this group and stop receiving emails from it, send an 
> email to [email protected].
> To post to this group, send email to [email protected].
> Visit this group at https://groups.google.com/group/sage-release.
> For more options, visit https://groups.google.com/d/optout.

-- 
You received this message because you are subscribed to the Google Groups 
"sage-release" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
To post to this group, send email to [email protected].
Visit this group at https://groups.google.com/group/sage-release.
For more options, visit https://groups.google.com/d/optout.

Reply via email to