Building on freebsd and hurd about luaposix HOT 6 CLOSED

Thanks.

I'm pretty sure I've fixed freebsd differently by being consistently more conservative about defining preprocessor constants to Lua.

I also took the #ifdef __GNU__ half of your Pgetcwd patch, but I don't understand why you need to define _GNU_SOURCE in lposix.c before config.h, which defines it anyway; and I don't see the need for an additional error message in the non GNU half of your Pgetcwd patch, since Hurd never sees that code anyway, right?

Please close this issue if current git HEAD works for you, or else let me know of any additional changes you need.

Cheers,
Gary

from luaposix.

On Sun, Aug 25, 2013 at 10:21:54PM -0700, Gary V. Vaughan wrote:

I also took the #ifdef __GNU__ half of your Pgetcwd patch, but I
don't understand why you need to define _GNU_SOURCE in lposix.c before
config.h, which defines it anyway; and I don't see the need for an
additional error message in the non GNU half of your Pgetcwd
patch, since Hurd never sees that code anyway, right?

I guess you are right for _GNU_SOURCE.
Check the return value for pathconf is not Hurd specific, but should be
done IMO (on Hurd it was always failing, but may fail on linux too
according to the manpage).

I've just uploaded a new version that disables the kfreebsd patch and
uses your commit for Hurd. It is called 29-4, let's see how it builds:

https://buildd.debian.org/status/package.php?p=lua-posix

Cheers

Enrico Tassi

from luaposix.

On Sun, Aug 25, 2013 at 10:21:54PM -0700, Gary V. Vaughan wrote:

Thanks.

I'm pretty sure I've fixed freebsd differently by being consistently more conservative about defining preprocessor constants to Lua.

I also took the #ifdef __GNU__ half of your Pgetcwd patch, but I don't understand why you need to define _GNU_SOURCE in lposix.c before config.h, which defines it anyway; and I don't see the need for an additional error message in the non GNU half of your Pgetcwd patch, since Hurd never sees that code anyway, right?

Please close this issue if current git HEAD works for you, or else let me know of any additional changes you need.

No, it did not work... I guess we were both wrong about _GNU_SOURCE ;-)

Ciao

Enrico Tassi

from luaposix.

Looks like 29-5 worked for you, but I can't find a record of what that patch contains.

Is the early _GNU_SOURCE definition required after all? Anything else?

Regarding checking pathconf result on non-GNU: rather than fall over if pathconf returns -1, we already just fall back to _POSIX_PATH_MAX, which seems reasonable to me.

Cheers,
Gary

from luaposix.

On Mon, Aug 26, 2013 at 07:14:52PM -0700, Gary V. Vaughan wrote:

Looks like 29-5 worked for you, but I can't find a record of what that
patch contains.

in 29-5 I restored the old patches, 29-4 failed on both freebsd and
hurd. Here the relevant logs.

https://buildd.debian.org/status/fetch.php?pkg=lua-posix&arch=hurd-i386&ver=29-4&stamp=1377510107
https://buildd.debian.org/status/fetch.php?pkg=lua-posix&arch=kfreebsd-amd64&ver=29-4&stamp=1377509665

Cheers

Enrico Tassi

from luaposix.

Hi Enrico, thanks for the logs.

It looks like your 29-4 patch was missing some of the portability changes I pushed to fix issue #108, since all of the undefined macros in the logs are wrapped with e.g #ifdef SA_NOCLDWAIT after my change sets in response to your report. At a minimum I expect you would need to include: e42d443 and 4b5b8ce, plus I just cleaned up all of the warnings from your 29-4 build logs with cd2e420, f9d490d, 3ae2025 and 1757152.

I don't know how well those will apply out of order, but if you can confirm that current master works on freebsd and hurd, I'll roll it up and release luaposix-30.

Thanks again!

from luaposix.