Supporting TAP tests with MSVC and Windows - Mailing list pgsql-hackers

From Michael Paquier
Subject Supporting TAP tests with MSVC and Windows
Date
Msg-id CAB7nPqTQwphkDfZP07w7yBnbFNDhW_JBAMyCFAkarE2VWg8irQ@mail.gmail.com
Whole thread Raw
Responses Re: Supporting TAP tests with MSVC and Windows
List pgsql-hackers
Hi all,
(Adding Peter and Heikki in CC for awareness)

Please find attached a WIP patch to support TAP tests with MSVC. The
tests can be kicked with this command:
vcregress tapcheck

There are a couple of things to note with this patch, and I would like
to have some input for some of those things:
1) I have tweaked some code paths to configure a node correctly, with
for example listen_addresses = 'localhost', or disabling local
settings in pg_hba.conf. Does that sound fine?
2) tapcheck does not check if IPC::Run is installed or not. Should we
add a check in the code for that? If yes, should we bypass the test or
fail?
3) Temporary installation needs to be done in the top directory,
actually out of src/, because Install.pm scans the contents of src/ to
fetch for example the .samples files and this leads to bad
interactions. Using this location has as well the advantage to install
the binaries for all the tests only once.
4) pg_rewind tests are disabled for now, I am waiting for the outcome
of 5519F169.8030406@gmx.net
5) ssl tests are disabled for now (haven't looked at that yet). They
should be tested if possible. Still need some investigation.
6) the command to access pg_regress is passed using an environment
variable TESTREGRESS.
7) TAP tests use sometimes top_builddir directly. I think that's ugly,
and if there are no objections I think that we should change it to use
a dedicated environment variable like TESTTOP or similar.
8) Some commands have needed some tweaks because option parsing on
Windows is... Well... sometimes broken. For example those things do
not work on Windows:
createdb foobar3 -E win1252
initdb PGDATA -X XLOGDIR
Note that modifying those commands does not change the test coverage.
9) @Peter: IPC::Run is not reliable when using h->results, by
switching to (h->full_results)[0] I was able to get results reliably
from a child process.
10) This patch needs to have IPC::Run installed. You need to install
it from CPAN, that's the same deal as for OSX.
11) Windows does not support symlink, so some tests of pg_basebackup
cannot be executed. I can live with this restriction.

I think that's all for now. I'll add this patch to the 2015-06 commit fest.
Regards,
--
Michael

Attachment

pgsql-hackers by date:

Previous
From: Michael Paquier
Date:
Subject: Re: Logical decoding (contrib/test_decoding) walsender broken in 9.5 master?
Next
From: Abhijit Menon-Sen
Date:
Subject: Re: What exactly is our CRC algorithm?