[tex-live] initial builds for tl'16

[Mojca Miklavec]

On 5 March 2016 at 09:13, Marc Baudoin wrote:
>
> As I already reported back in September 2015 on tlbuild (message
> "LuaJIT segfaults on NetBSD"), everything builds fine on NetBSD
> but the test suite fails for luajit:
>
> There has been no progress since then.

There has been progress. We diagnosed the problem and Luigi provided
an ad-hoc patch; the only problem is that he only fixed it for OpenBSD
and Solaris (and even that is somewhat inconsistent compared to other
platforms).

See http://tug.org/svn/texlive?view=revision&revision=39893

Here's my (off-list) response (I can provide more context if needed;
but the major problem is that source/libs/luajit/m4/lj-system.m4
treats any platform that's not in a list Darwin/Linux, now also
SunOS/FreeBSD as "other non-POSIX platform" which leads to problems):


On 1 March 2016 at 10:38, Mojca Miklavec wrote:
> On 29 February 2016 at 17:11, luigi scarso wrote:
>>
>> in source/libs/luajit/m4/lj-system.m4
>> I have added
>> [SunOS | FreeBSD], [],
>
> What about NetBSD, OpenBSD, CYGWIN and potentially others?
>
> And just curious: why not be consistent with other platforms and
> either not defining anything for any of them or perhaps do something
> like:
>
>     [SunOS], [echo '-DLUAJIT_OS=LUAJIT_OS_POSIX' >>native_flags],
>     [FreeBSD|OpenBSD|NetBSD], [echo '-DLUAJIT_OS=LUAJIT_OS_BSD' >>native_flags],
>
>> so that for the next luatex 0.89.3 they should be ok, but I think that for
>> TeXLive
>> lj-system.m4 should have a better fix.
>
> Yes, it would be great to have a better fix. And it would be desirable
> to test the behaviour on Cygwin. Judging from the code I would bet
> that it would have problems (or maybe not because it has some backup
> from the underlying Windows system).
>
> Can someone also come up with a simple additional test for luajittex
> to be included in TeX Live?

Mojca