gcc48: backport of cmath fixes

[rmottola (Riccardo)]

I backported this fix:

​https://gcc.gnu.org/viewcvs/gcc?view=revision&revision=244231

of which the comments are: ​https://gcc.gnu.org/bugzilla/show_bug.cgi?id=79017

Contrary to the original bug, it is not PPC only, but x86 too

to gcc48. It was pointed out by Ken, it is needed to compile TenFourFox. The fix works. I have run the compiler test suite before and after and it has no adverse effects.

I opened an upstream bug: ​https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87635

I don't have faith in a new release of an unsupported compiler. I'd appreciate the inclusion in MacPorts as patches (and if 4.8.6 comes out.. we can remove them, can't we)

I can also consider backporting them to gcc6, now that I am "expert".

[mf2k (Frank Schima)]

Thanks. Note that you are more likely to get this committed, and faster, if you submit a GitHub pull request.

[kencu (Ken)]

I would support including this fix, if agreeable to others, and appreciate the work Riccardo has done on it to prove it's functional worth.

[ryandesign (Ryan Carsten Schmidt)]

I see you filed #57656 to backport these changes to gcc6. Are the other gcc versions, such as gcc49 and gcc5, affected as well?

[rmottola (Riccardo)]

I udnerstand that all versions up to gcc7 when it was fixed are affected. I didn't think gcc5 was used at all in ports? With some fuzz, using either the gcc48 or the gcc6 patches will probably apply to gcc49 and gcc5

[kencu (Ken)]

We decided to fix this in the legacysupport workings instead, and Michael did that about six months ago. If that doesn't work, we might need to revisit this idea.

[rmottola (Riccardo)]

I think we will never do that, relying on LegacySupport, right? Then it is better to close this bug.

[rmottola (Riccardo)]

I am unable to close this. The patches are there, but things were worked around with LegacySupport. ArcticFox now compiles with newer compilers, so this is less important now.