New Windows animal, mostly ecpg trouble (was: Crash with old Windows on new CPU)

From: Christian Ullrich <chris(at)chrullrich(dot)net>
To: Alvaro Herrera <alvherre(at)2ndquadrant(dot)com>, Magnus Hagander <magnus(at)hagander(dot)net>
Cc: Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>, Robert Haas <robertmhaas(at)gmail(dot)com>, "pgsql-hackers(at)postgresql(dot)org" <pgsql-hackers(at)postgresql(dot)org>, Michael Meskes <meskes(at)postgresql(dot)org>
Subject: New Windows animal, mostly ecpg trouble (was: Crash with old Windows on new CPU)
Date: 2016-03-10 15:51:17
Message-ID: 56E197F5.6000704@chrullrich.net
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

* Alvaro Herrera wrote:

> Magnus Hagander wrote:

>> On Wed, Mar 9, 2016 at 4:36 PM, Christian Ullrich <chris(at)chrullrich(dot)net>
>> wrote:

>>> And apparently not a single one with VS 2013. OK, I'll see what I can do
>>> about setting some up soonish, at least with (server) 2008 and (client) 7.
>>> FWIW, I have a local build of 9.5.1 with this patch in daily use on 2008
>>> now, with no complaints.
>>
>> Having that added to the buildfarm would definitely be very useful!
>
> Yeah, ideally we would have one buildfarm member for each MSVC compiler
> supported by each Windows version. AFAICS we only have
>
> Windows 8 8.1 Pro Visual Studio 2012 x86_64
> Windows Server 2008 R2 (64bit) Visual C++ 2005 AMD64
> Windows Server 2003 R2 5.2.3790 MSVC 2005 Express 8.0.50727.42 x86
> Vista Ultimate 6.0.6000 MSVC 2005 Pro 8.0.50727.867 x86
> Windows XP-PRO SP3 MSVC++ 2008 Express i686

Add to that Windows 7 with Visual Studio 2013 (x64), named woodlouse,
and there will be another one sometime over the weekend. Setting up
Windows animals is a major pain; no wonder so few people run any.

I spent some hours getting rid of random linker errors in ecpg
(kernel32.lib not found, msvcrt.lib not found, or unresolved symbols
that should be in one of the two); they finally went away when I stopped
defining the build environment in build-farm.conf and ran vcvarsall.bat
instead (this was probably pilot error; I think I had a mix of x86 and
x64 paths there). At that point, I managed one successful run on HEAD.
Things started failing in ecpg again after that, with crashes of the
test programs for the thread-thread and thread-prep test cases,
apparently somewhere in setlocale(). I had to --skip-steps=ecpg-check
because the crashes cause blocking UI.

The other troublemaker is plperl; it produces a wall of compiler errors
which I cleverly did not copy, most are about something not being in a
formal parameter list. Some preprocessor trouble, I suspect. This is
with both ActiveState and Strawberry Perl.

I did not try Tcl, and plpython3 apparently builds, but I cannot see any
tests being run.

--
Christian

In response to

Browse pgsql-hackers by date

  From Date Subject
Next Message Alvaro Herrera 2016-03-10 15:51:56 Re: create opclass documentation outdated
Previous Message Stephen Frost 2016-03-10 15:40:36 Re: Add generate_series(date,date) and generate_series(date,date,integer)