Bump to 6.8.0.123 (still the 2019-10 branch), clean things up a bit more.

In D24988#550221, @pkubaj wrote:

Looking at the diff between pkg-plist and pkg-plist.powerpc, I noticed I missed this at the end of pkg-plist.powerpc:

@postexec %%PREFIX%%/bin/cert-sync %%LOCALBASE%%/share/certs/ca-root-nss.crt
@postunexec rm -rf %%LOCALBASE%%/share/.mono

Please add this.

Noting here that as of today, all patches to 6.8.0.105 have been upstreamed
https://github.com/mono/boringssl/pull/23
https://github.com/mono/mono/pull/19234

In D23300#529892, @pkubaj wrote:

It doesn't apply after r526962:

Apologies for the apparent radio silence; I've actually been working on some high priority issues in upstream and trying to sort out inotify. The patch issue with ppc64 was actually just an SVN hiccup. Upstream patch is correct and will be in next release. This should be ready to go for ppc64 and aarch64.
@greg_unrelenting.technology - your patch should ONLY need to touch external/boringssl/crypto/cpu-aarch64-linux.c at this point, but I can't test here. If that's the case, can you please submit a PR with it to https://github.com/mono/boringssl ?

• Reorder Makefile and expand to svn diff -x -U999999 as requested by @mat
• Fixed typo in files/patch-powerpc64; this now matches powerpc*-*-freebsd* as in upstream
• Correctly mark BROKEN for QEMU_EMULATING; Roslyn is very picky about POSIX and does not like qemu
• Fix .if ${ARCH} PLIST handling by switching to pre/post.mk; accurate plists are still needed for ppc64 ELFv1, arm*, and arm64* This is based on https://reviews.freebsd.org/D22451
• Remove "EXPERIMENTAL" from Ninja; this is well-tested and perfectly safe. It will remain non-default because of the number of dependencies it pulls in.
• Remove 'TZ' from MAKE_ENV; this actually hasn't been necessary for a while
• Correct branch comment; 6.8.0.105 is actually branch 2020-02

In D23300#518554, @pkubaj wrote:

On powerpc64, I'm still getting:

configure: error: unknown target

This is because in configure.ac, there's neither powerpc64-*-freebsd, nor powerpc*-*-freebsd:

! In D23300#518091, @driesm.michiels_gmail.com wrote:

This last version update has some plist issues for me as well (amd64 on STABLE12). Please check with Poudriere for QA before updating the diff. If possible. ;-).

• Update to 6.8.0.105
• Includes fix for powerpc64 building (patch has been upstreamed)
• Includes @greg_unrelenting.technology 's patch partially; testing is required here. All of the Mono changes were already done upstream except for the boringssl portion. Some were done slightly differently (e.g. REDZONE is for all !APPLE) and revisions were necessary to boehm being deprecated/externalized.

This makes perfect sense to me. Some ports require Mono as a runtime (because they get called as 'mono blah.exe') but some ports don't, and Mono is a pretty hefty build.

@pkubaj can you please drop this attachment into files/ and re-test? Thanks!

@pkubaj Amusingly, those plist errors are fully expected - ppc ISA produces different binary sets than x86. I forgot to note the change to Mono's makefiles to turn on ppc64; that's definitely on me. But honestly, I'm far more surprised it built at all. (I only have qemu where ppc64 is "no.") Can you please try without --disable-btls? I do expect failure there, but, hey! Maybe we'll be surprised, right?

Many thanks to @mikael for attempting an aarch64 hardware build today. This successfully confirmed that it cannot be built under qemu due to a qemu issue. Unfortunately I don't know of any way to say 'for aarch64 but not if it's qemu.' The aarch64 build also encountered a long hang in the Roslyn (MCS) compiler which is a commonly known problem with Roslyn itself, also commonly seen under musl libc Linux. Specifically spinning locks; several issues have been opened upstream and some are still unfixed. The build may need a LONG total time (potentially many hours) if it successfully advances past these hangs, but it needs a solid 4 hours of spinning gears to see if it can move past it.

Note that I generated from an svndiff and a bit of fumbling in the dark, so there probably are some minor glitches here.
The comments identifying the branch and the date/hash of monolite are important for ease of maintenance. This port will be pretty fast moving.