Darcy Buskermolen wrote:
> On Saturday 05 August 2006 08:52, Stefan Kaltenbrunner wrote:
>> cbbrowne at ca.afilias.info wrote:
>>>> On Friday 04 August 2006 15:30, Christopher Browne wrote:
>>>>> Thanks to all those that have been putting in effort to knock the dust
>>>>> off of the 1.2 code.
>>>>>
>>>>> It has been a reasonably busy week of knocking off little problems
>>>>> here and there:
>>>>>
>>>>> - A fix to let 1.2 "play well" with some restructuring of #includes in
>>>>> 8.2
>>>>> - Added a new regression test that tests inheritance
>>>>> - People turned up GCC to -Wall (generate *ALL* warnings), and have
>>>>> stomped
>>>>>   virtually all warnings
>>>>> - Cleaning up of Makefiles
>>>>> - Cleaning up of SGML tagging problems in documentation
>>>>> - Added a table with extra types (inet, numeric, float, various
>>>>> geometric
>>>>> types) to "test1" regression test
>>>>>
>>>>> Darcy discovered a bug in 8.2 with domains with checks, and it appears
>>>>> a fix should be in place forthwith; the plan seems to be that in view
>>>>> of holidays in Canada, we'll revisit 8.2-based testing on Tuesday, and
>>>>> see about RC3 on Wednesday.
>>>> A quick run through test/* on my 8.2 dev box come up clean, I'll see
>>>> about setting up a several hundred gig test before I leave for the day,
>>>> but so far
>>>> I think we are green light for an RC3.  the only outstanding issue I
>>>> know of
>>>> at this point in time is windows builds.
>>> Thinking about it, Windows may be less of an issue than was thought,
>>> although there may be some work that *ought* to be done...
>>>
>>> The trouble Mastermind hit was that the process of building for Windows
>>> seems underdocumented, not that we broke it recently.
>>>
>>> My suspicion is that what's missing is some feedback as to configure
>>> changes that need to be done on Windows.  I expect that people doing
>>> builds on Windows have been making some changes and specifying much more
>>> in terms of configure parameters than are needful on Unix-like systems.
>> well I found some time to play a bit more with that and after some
>> hacking on both the makefiles and the configure script we now have what
>> maybe is a real issue:
>>
>> dllwrap -o slony1_funcs.dll --def slony1_funcs.def slony1_funcs.o
>> win32ver.o
>> /home/pgbuild/devel/pginst81/lib/postgresql/pgxs/src/utils/dllinit.o
>> -L/home/pgbuild/devel/pginst81/lib -lpostgres
>> Info: resolving _DateStyle by linking to __imp__DateStyle (auto-import)
>> fu000001.o(.idata$3+0xc): undefined reference to `libpostgres_a_iname'
>> fu000002.o(.idata$3+0xc): undefined reference to `libpostgres_a_iname'
>> fu000003.o(.idata$3+0xc): undefined reference to `libpostgres_a_iname'
>> fu000004.o(.idata$3+0xc): undefined reference to `libpostgres_a_iname'
>> fu000005.o(.idata$3+0xc): undefined reference to `libpostgres_a_iname'
>> fu000006.o(.idata$3+0xc): more undefined references to
>> `libpostgres_a_iname' follow
>> nmth000000.o(.idata$4+0x0): undefined reference to `_nm__DateStyle'
>> collect2: ld returned 1 exit status
>> c:\MinGW\bin\dllwrap.exe: c:\MinGW\bin\gcc exited with status 1
>> make[2]: *** [slony1_funcs.dll] Error 1
>> make[2]: Leaving directory `/home/pgbuild/devel/slony1-engine/src/backend'
>> make[1]: *** [all] Error 2
>> make[1]: Leaving directory `/home/pgbuild/devel/slony1-engine/src'
>> make: *** [all] Error 2
>>
>> that might point to an issue with the recent datestyle fixes ...
> 
> I've commited a fix for this.  Looks like the direct manipulation of GUC vars 
> is not possible under win32, instead we must use the GUC helper functions.

with that fix in place slony1 now compiles and passes a number of
regression tests(testddl,test1,testdatestyle) on my development box - so
it looks like we can go on with RC3 and leave usability improvements(in
terms of buildsystem fixes) for a possible 1.2.x release.


Stefan