[tlbuild] asy2.49 build

[Mojca Miklavec]

On Thu, 4 Apr 2019 at 00:50, Karl Berry wrote:
>
> I updated the asy runtime files with 2.49
> (and the x86_64-darwin binary, thanks Dick),
> although the usefulness of TL continuing to distribute any asy binaries
> remains unclear to me.
>
> At any rate, I think we are just about ready to start the final push
> toward the release ... except for asy. If you can't build a useful
> binary, I suggest just dropping it. I don't much care for the idea of
> indefinite delay while all these OpenGL portability problems are batted
> around. --best, karl.

My summary:

(1) For Solaris I successfully built asymptote without OpenGL support
(it has never been there and nobody ever complained about it). Just
for fun I tried to compile with defaults, but it's enough to confuse
gmake to not even be able to finish the build (looks like a broken
Makefile). I can come back to this after TL release.

(2) On FreeBSD I have the same issue with GC as Dick reported for
macOS, but reverting to an older GC didn't really help (it didn't help
me on macOS either). Apparently the code doesn't make clang happy
(neither on FreeBSD nor on macOS). It would be really helpful if
someone could help me come up with some minimal example that could be
reproduced outside of asymptote, so that we could at least report it
upstream in a proper way. The author of GC responded:
   "A lot of C++ changes were made since 7.6.0. Could you please try
other 7.6.x versions? Is 7.6.12 working? If not, which is latest 7.6.x
working?"
I cannot answer that question since for me reverting didn't have any
effect. Dick?

(3) I seem to have serious performance issues while compiling on Mac
OS X 10.6 that haven't been there a few days ago (it gets stuck at
"checking for fork..." {5 minute break} "yes", and then the same for
any single functionality check). Letting it compile over night didn't
help. I'll try to figure out if there's anything I can do about that
on the machine itself, or switch to another one; "top" says
"softwareupda 43.3 3124 hrs" (spending 43% of CPU running some strange
"softwareupdate" for the last three thousand hours???? Apple doesn't
even release any updates any longer.) Meanwhile the full TeX Live
build can be performed without a hiccup (yes, it takes almost 4 hours,
but that has always been the case).

"ps -a" shows me as if I ran the asy build twice (once the fork checks start):

  PID TTY           TIME CMD
24319 ttys000    0:00.25 -bash
26239 ttys000    0:00.02 /bin/bash ./do_build_asy.sh
27801 ttys000    0:00.33 make
30389 ttys000    0:01.98 /bin/sh ./configure --enable-texlive-build
--enable-gsl --enable-fftw CXXFLAGS=-ansi -O3 -isysroot
/Developer/SDKs/MacOSX10.6.sdk -mmacosx-version-min
31564 ttys000    0:00.01 /bin/sh ./configure --enable-texlive-build
--enable-gsl --enable-fftw CXXFLAGS=-ansi -O3 -isysroot
/Developer/SDKs/MacOSX10.6.sdk -mmacosx-version-min
31565 ttys000    0:00.11 /opt/local/libexec/llvm-7.0/bin/clang++
-stdlib=libc++ -o conftest -ansi -O3 -isysroot
/Developer/SDKs/MacOSX10.6.sdk -mmacosx-version-min=10.6 -I/Use
31567 ttys000    0:08.74 /opt/local/libexec/llvm-7.0/bin/clang -cc1
-triple x86_64-apple-macosx10.6.0 -Wdeprecated-objc-isa-usage
-Werror=deprecated-objc-isa-usage -emit-obj -

and after a while I figured out that I can actually run "Ctrl+C" in
the terminal where I started the build and the build continues at
normal speed. Sometimes. Sometimes it breaks. So far I seem to be
unable to compile on darwinlegacy.

(4) I had issues compiling on a recent version of macOS (switching to
an older version did not help), but since I'm not the one officially
compiling it for TL, this can safely be ignored. I also had issues
getting it to work properly for MacPorts, but that's another thing
that has nothing to do with TeX Live release.


Mojca