On Sat, Nov 09, 2019 at 03:55:18PM +0100, Landry Breuil wrote:
> Hi,
>
> sqlite3-tcl is a bit outdated compared to 'main' sqlite3, so im
> pondering moving back to the previous situation of having them as a
> single port with multipackages (which was the case before sqlite3 got
> imported into base) - and while here i hacked my way around to be able
> to run the tcl test suite. Right now its still running (and seems hung)
> but that allows to run those natively for each updates.
>
> - we need to switch to the -src tarball which is the only one containing
> tests
> - that requires horrible hacks to generate configure in the autoconf/tea
> subdir, + generate tclsqlite3.c wrapper.
> - that means rerunning configure/building in do-test as the WRKDIST dir
> changes
>
> so its a gross hack, more an interesting exercise in ports wrestling,
> but finally having sqlite tests. even if it doesnt build with the same
> options as the main sqlite3 port.. some of the SQLITE_ENABLE_XX choices
> look rather arbitrary.
And the resulting libsqlite3 tcl library is actually completely broken,
so yeah its a gross hack. But that said, i managed to run the complete
tcl testsuite, which resulted after smth like 7h in not so bad results:
SQLite 2019-10-10 20:19:45
18db032d058f1436ce3dea84081f4ee5a0f2259ad97301d43c426bc7f3dfalt1
13 errors out of 3444413 tests on c64.proxmox2 OpenBSD 64-bit little-endian
!Failures on these tests: backup2-10 loadext-2.1 loadext-2.2 loadext-2.3
loadext-3.2 sort3-3 memsubsys1.backup2-10 memsubsys2.backup2-10
no_mutex_try.backup2-10 journaltest.backup2-10 inmemory_journal.backup2-10
prepare.backup2-10 mm-backup2-10
(and i think the failures are not actual failures)
so still, this *might* be interesting to be able to run those tests somehow, so
that one updating sqlite3 can be a bit more comfortable with testing.
Landry
