testing

[Olly Betts]

On Mon, Dec 12, 2016 at 05:00:36PM +0000, Eric Lindblad wrote:
> If the subject of xapian-core apitests on MSYS/MINGW and MSYS2/MINGW is taken
> up it might be noted that the compiler version could be recorded, probably
> gcc 5.3.0 or newer, that closed pipes might be an issue, and also allocating
> for the location of xapian-tcpsrv.
> 
> Not from my own tests but another man informed there was some problems with
> [MSYS2] python bindings.

That's unhelpfully vague, but perhaps this patch helps with "closed
pipes"?

https://github.com/ojwb/xapian/commit/43af24c4a39819463621e9f8494cf7fdd30736e4.patch

If so, let me know and I'll apply it to git master.

> I am not knowledgeable of the packages and/or files utilised with MSYS/MINGW
> and/or MSYS2/MINGW to resolve a hostname, however, maybe a xapian-core
> compile and run of 'make check' under MSYS/MINGW would provide more output
> than (as below) under MSYS2/MINGW, which merely printed "FAIL: apitest.exe".

You ought to see the failing tests - if you aren't, I'm not sure why not.

Perhaps try:

make VERBOSE=1

If not that, perhaps you've somehow managed to enable automake's TAP test
harness, in which case the actual output will probably be in "*.log"
files.

> If there is any usefulness in MSYS/MINGW output in getting MSYS2/MINGW to
> pass the apitest I can not say, neither will forward I an assessment as to
> whether in this direction old stable (1.2.24) might be looked at before
> current (1.4.1) as pertains to MSYS/MINGW and/or MSYS2/MINGW.

1.2.x is now firmly in maintenance mode.  If people provide clean patches
for portability issues, they're very likely to get applied, but my focus
is very much on 1.4.x and git master.

Cheers,
    Olly