Missing or incomplete (Free)BSD support

An alternative solution to the two attached files were added in [3174]

bash placement portability in [3176]

What is the version identifier used on FreeBSD, and if there are both BSD and FreeBSD, is BSD generally general enough to use?

Another question; which version of GDC?

I'm not deeply familiar with the other BSDs, but my guess is that they're generally compatible for most purposes relevant to tango. I'm on FreeBSD-7.0-PRERELEASE, and running the version of gdc in ports, which reports itself as: {{{gdc (GCC) 4.1.3 20071105 (prerelease (gdc 0.24, using dmd 1.020)) Copyright (C) 2006 Free Software Foundation, Inc. This is free software; see the source for copying conditions. There is NO warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.}}}

My gdc version defines version "freebsd" (note case) on my system, making "version(freebsd)" the appropriate conditional.

Preliminary support in [3187] and [3188].

(In [3191]) FreeBSD definitions, thanks Ralith. Refs #896

(In [3193]) support for freebsd, thanks Ralith, refs #896

I'm a bit confused about these diff files. The first few are all ads, but the last one alters stuff in the freebsd blocks and also alters things in general Tango source files. The first issue leads me to wonder if the adds were actually done based on the freebsd headers or if they were simply a copy of the linux declarations. If the latter, why was this done? As for the second issue, I'd like to believe that if freebsd stuff is defined properly in the Posix modules then no other changes should be necessary anywhere, yet I see a bunch of new freebsd-specific versioning in the Thread and Socket modules, among others.

Many of the ads were in cases where there were no 'default' declarations; without them, the relevant functions would simply not be declared on FreeBSD, and there sometimes even existed a static assert(false) that would prevent compilation from continuing without such a duplication. The edits in Thread.d are to accommodate the fact that FreeBSD's pthread cleanup funcs are actually funcs, and thus must be used as such, whereas Linux (and darwin?) used macros that were approximated using structs. The linux/darwin pthread cleanup stuff could probably be modified to provide a unified function interface that would remove the need for this, but that was significantly beyond my task. The Socket.d edits are to accommodate for minor differences in the structs used by FreeBSD vs. other systems as declared in that file. I agree that these structs should be declared elsewhere, but I opted for simply adding to what already existed, instead of reorganizing.

Mostly working and relative to latest SVN. Includes fully-working (although only double-accurate) math funcs. Testing, determination of issues, and fixes for them needed!

Hmm, seems like the example hangs on Linux too - albeit not so often if I exchange Stdout with log.Trace.

Ralith; an additional thing you can do, go to lib/ and run

./unittest.sh --run-all gdc

then run the resulting runSomething binary. There should be about 60 modules/tests listed.

Handing over to Sean as he wants to review.

Have you tried locking the appropriate mutex before issuing the notify?

I'll accept patches for the C / Posix headers, but I'd prefer if they affect only those files. Please follow the example of how the code is currently formatted (version blocks per OS) and submit a patch for what you'd like added. Or simply submit text files with the additions and I can put them in. If any Tango user code is to be altered I'd prefer that be done in a separate ticket with an explanation for why it's necessary.

WrongProcedureException! I agree with the sentiment that C/Posix patch should be in a different ticket from any user library ticket, but we still need a !FreeBSD support ticket and I think this one should be it.

So, to move on, create a ticket for C/Posix patches, and a ticket for each of any other issues that may exist, and link those necessary for BSD support to this ticket such that this one can be closed when this support is stable'ish.

A cleaner and somewhat improved patch for the C portions of tango has been uploaded to #1225

A cleanup patch for Thread.d (arguably worth applying even without FreeBSD, to make it sync closer with tango/stdc/posix/thread.d) has been uploaded to #1227

A patch declaring needed system-specific alongside those for Linux, Darwin, and Windows has been uploaded to #1228

A patch disabling tango.math.Math.log2, which depends on a C/POSIX function unavailable on FreeBSD-STABLE, has been uploaded to #1229

#1229 closed; an implementation of log2 has been included in the second update to the patch in #1225.