
Hi Karel, A few months ago we reverted a cleanup [1] to PosixSource.h due to breakage on Solaris. I believe I have found [2] a build log from a failing build, but sadly it's far from clear what is going on. Do you suppose you could send us your /usr/include/sys/feature_tests.h? I suspect this will shed some light on what has gone wrong. For the record, I've opened #12624 to track this issue. Cheers, - Ben [1] https://git.haskell.org/ghc.git/commitdiff/cac3fb06f4b282eee21159c364c4d08e8... [2] http://haskell.inf.elte.hu/builders/solaris-amd64-head/573/10.html

Hi Ben, sending Oracle copyrighted header file to public mailing list is no-go for me, I hope you understand it. :-) Anyway, the problem with the patch[1] is simple: Oracle's Solaris 11 supports max: #define _POSIX_C_SOURCE 200112L #define _XOPEN_SOURCE 600 the patch[1] assigned values: #define _POSIX_C_SOURCE 200809L #define _XOPEN_SOURCE 700 which are not recognized by Oracle's feature_tests.h hence the complain and #error. I've verified this fact on both Solaris 11.2 and Solaris 11.3. On the other hand Illumos derivates should be OK with the definition of patch[1] as Illumos sometime in its history added handling for XPG7[2]. Basically if you look into history of Illumos feature_tests.h and you get the first commit[3], then you basically will see what's in Solaris 11.2/11.3 minus few really unrelated changes done in Solaris. I hope it's more clear now, what's happened with[1] on Solaris. Thanks! Karel [1] https://git.haskell.org/ghc.git/commitdiff/cac3fb06f4b282eee21159c364c4d08e8... [2]: https://github.com/illumos/illumos-gate/blob/master/usr/src/uts/common/sys/f... [3]: https://raw.githubusercontent.com/illumos/illumos-gate/7c478bd95313f5f23a4c9... On 09/26/16 09:17 AM, Ben Gamari wrote:
Hi Karel,
A few months ago we reverted a cleanup [1] to PosixSource.h due to breakage on Solaris. I believe I have found [2] a build log from a failing build, but sadly it's far from clear what is going on. Do you suppose you could send us your /usr/include/sys/feature_tests.h? I suspect this will shed some light on what has gone wrong.
For the record, I've opened #12624 to track this issue.
Cheers,
- Ben
[1] https://git.haskell.org/ghc.git/commitdiff/cac3fb06f4b282eee21159c364c4d08e8... [2] http://haskell.inf.elte.hu/builders/solaris-amd64-head/573/10.html
participants (2)
-
Ben Gamari
-
Karel Gardas