devel/valgrind-devel: update to Paul Floyd's FreeBSD valgrind branch

[kevans]

Summary

devel/valgrind-devel: update to Paul Floyd's FreeBSD valgrind branch

This branch rebases us forward to Valgrind 3.17.0, incredibly current as
of today. It also fixes a vast array of known issues, making valgrind-devel
actually functional on 13.0-CURRENT and fixing most of the regression tests
on clang/amd64.

All patches previously in the port have been incorporated.

There are still some open issues remaining:
- https://github.com/paulfloyd/freebsd_valgrind/issues

This port is believed to be largely functional enough to get wider testing. Many,
many thanks to the excellent work of Paul Floyd to modernize our valgrind port.

Test Plan

Q/A:

• portlint (some pre-existing issues to address separately, some addressed here)
• testport (11.3/amd64, 12.1/i386, -CURRENT/amd64)

This is largely untested (at runtime) on 11.x, it would be good to get some testing there before commit as well. I've briefly tested it on -CURRENT/amd64 and Paul develops on 12.1/amd64.

[pjfloyd_wanadoo.fr]

I had a go at building, replacing valgrind-dev Makefile and distinfo. I got

fetch: https://codeload.github.com/paulfloyd/freebsd_valgrind/tar.gz/0511105?dummy=/paulfloyd-freebsd_valgrind-freebsd-3.17.0.g20200625-0511105_GH0.tar.gz: size of remote file is not known
paulfloyd-freebsd_valgrind-freebsd-3.17.0.g202          14 MB   18 kBps 13m18s
=> Fetched file size mismatch (expected 15485736, actual 15408039)

The pkg-plist diffs look OK from a Valgrind pov.

[kevans]

Whoops- I'll update this to the latest commit and upload a new diff soonish.

[pjfloyd_wanadoo.fr]

Looks OK now. make && make install worked, running the regression suite on the installed Valgrind gave me expected redults.

[emaste]

LGTM

[pjfloyd_wanadoo.fr]

Accepted

[markj]

JFYI, I submitted a status update referencing the port update: https://github.com/freebsd/freebsd-quarterly/pull/212

[pjfloyd_wanadoo.fr]

I've put much more detailed regtest failure analysis here

regtest status

In summary:

• On FreeBSD 11, there's one testcase that hangs (one that uses a pthread condition variable that has been trashed). There are about 10 extra failures wrt 12.1, 3 need further investigation and 7 look harmless
• On FreeBSD 13 I get about 50 more failures. Many of these are issues with clang 10 and Valgrind reading debuginfo to get to variable names and source file/line.