boost: 1.59.0

[dplarson]

This commit also removes the boost-python patch as the issue was fixed upstream (see boostorg/build#78 for more info).

[tdsmith]

As is customary for boost releases, this breaks a bunch of builds; my lab workstation is building them and I'll report back when it's done.

[tdsmith]

librevenge is broken and it has a lot of dependencies. Looks from the bot like this isn't ABI-compatible, either.

• apngasm https://gist.github.com/b303fb8cc2935af07c80; upstream; fixed in 3.1.6 (PR)
• avro-cpp https://gist.github.com/03736608223d42f45ab1; upstream; PR
• cppdom https://gist.github.com/1b809c96ad31fb95bd16; upstream; fixed by boneyarding in a2a72b4
• ledger https://gist.github.com/258c90fcde8c801287ae; upstream; fixed by applying Python: Fix condition for make_setter overload. boostorg/python#40
• quantlib https://gist.github.com/63f8844069d07cc0b3ed; upstream; fixed in 1.6.1 (PR)
• librevenge https://gist.github.com/6a23ce70aec41aa0d615; upstream, fix, PR
  libcdr Error: No logs. (librevenge)
  libetonyek Error: No logs. (librevenge)
  libfreehand Error: No logs. (librevenge)
  libmspub Error: No logs. (librevenge)
  libodfgen Error: No logs. (librevenge)
  libpagemaker Error: No logs. (librevenge)
  libvisio Error: No logs. (librevenge)
  libwps Error: No logs. (librevenge)
  writerperfect Error: No logs. (librevenge)

irrelevant failures:
pktanon expected replacement of "#include "Timer.h"\n" with "#include "Timer.h"\n#include "unistd.h"\n" not Boost-related; see #43084
supersonic https://gist.github.com/8eeaabcd0ce25c1f8f55 not Boost-related; see #43083

[dplarson]

Thanks for checking on this @tdsmith. Is librevenge the only broken formula? Or just the first so far?

For what it's worth, I'm okay with delaying the upgrade until the other packages have time to make revisions. I originally submitted the PR while I was checking whether boost 1.59.0 fixed an issue I had with another (non-Homebrew) package.

[tdsmith]

No, all of the formulas listed with build logs are broken in their own, special ways. (Or possibly the same way; I haven't investigated beyond trying to build them.)

[ghost]

I can't see the test results for some reason--is there a list of pending tasks or dependent packages which need fixing? I can investigate fixes for some if I know where to look.

[tdsmith]

Just my comment earlier on this thread. It looks like there's a patch for avro-cpp linked from their bug tracker that we could apply.

[afh]

@tdsmith Thanks for reporting the issue on ledger to the mailing list, we (the people contributing on ledger) are working on a new release that addresses the issue.

I'll make sure to update this issue and the the ledger homebrew formula accordingly.

[afh]

Issue #39 in boost python seems to be the cause for the broken ledger build.

[tdsmith]

Okay, we'll apply the patch from boostorg/python#40 to boost-python.

[tdsmith]

The only project left is cppdom, which hasn't seen a commit in three years and didn't respond to my month-old bug report, so I think it's headed for the boneyard. We should be able to land this this week.

[afh]

@tdsmith I verified that with boostorg/python#40 applied ledger builds just fine.

[dplarson]

Looks like after #45564, boost 1.59.0 was merged into master and so this issue can be closed.