[Debian-science-sagemath] arb: headers with generic names

[Jerome BENOIT]

On 11/09/18 08:56, Julien Puydt wrote:
> Hi,
> 
> On 11/09/2018 05:25, Ximin Luo wrote:
>> Jerome BENOIT:
>>> On 10/09/18 19:13, Erik Bray wrote:
>>>> On Mon, Sep 10, 2018 at 7:12 AM Ximin Luo <infinity0 at debian.org> wrote:
>>>>>
>>>>> Julien Puydt:
>>>>>> [..]
>>>>>>
>>>>>> Upstream, then, as there's nothing I can do about it in Debian...
>>>>>>
>>>>>
>>>>> Well, we can install on-purpose into a Debian-specific directory like /usr/include/arb instead of /usr/include. Then all the packages that uses arb will have to add -I /usr/include/arb to their compiler flags, or #include <arb/xxx.h> instead of #include <xxx.h>.
>>>>>
>>>>> In fact I swear we were doing this already, and that we were happy to maintain this patch from upstream because what upstream does is so ridiculous. Did this get reverted or something?
>>>>>
>>>>> Of course the best solution is to have upstream adopt best practises, but here the situation IMO is too ridiculous to tolerate, we should just fix it in Debian for now, and keep maintaining the patch - which should not be that bad, just mechanical renames.
>>>>
>>>> Rather than patching the sources, wouldn't it make more sense to just
>>>> add `-I/usr/include/arb` where necessary?  This can go in a .pc file
>>>> for arb.
>>>
>>> A .pc file would be very fine, but arb is at the end of a chain: gmp, mpfr, flint, arb
>>> gmp, mpfr, and flint do not come with a .pc
>>>
>>> For this reason, I guess arb must follows its predecessors.
>>> gmp and mpfr have basically one header, flint a bunch of them gathered in a folder.
>>> For this respect arb is more like flint: its headers may be gathered in a folder.
>>>
>>> Otherwise, I think that submitted a patch to the upstream author is the best approach.
>>
>> Putting them in a separate folder solves the problem of co-installation, but does not solve the problem of potentially-conflicting "#include <>" directives. To solve this problem the sources (and anyone depending on them) do have to be patched, ideally by upstream.
>>
> 
> Yes, flint-arb upstream should break things to make them more correct.
> 
> Adding .pc files to the various other upstreams would be a nice feature though.

Of course, but this must be done by the upstream maintainer themselves.
Note that most software using GMP comes with their own GMP macro (see mpfr).

Whatever, gmp and mpfr maintainers could do it since a very long time,
so I guess that they do not want to do so. On the other hand, they make
the effort to gather their respective header material into one header.

Jerome

> 
> jpuydt on irc.debian.org
> 
> _______________________________________________
> Debian-science-sagemath mailing list
> Debian-science-sagemath at alioth-lists.debian.net
> https://alioth-lists.debian.net/cgi-bin/mailman/listinfo/debian-science-sagemath

-- 
Jerome BENOIT | calculus+at-rezozer^dot*net
https://qa.debian.org/developer.php?login=calculus@rezozer.net
AE28 AE15 710D FF1D 87E5  A762 3F92 19A6 7F36 C68B

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 1528 bytes
Desc: OpenPGP digital signature
URL: <http://alioth-lists.debian.net/pipermail/debian-science-sagemath/attachments/20180911/70775591/attachment-0001.sig>