Re: ActivePerl build failures

On 22/02/2012 10:45, Tim Bunce wrote:
> On Wed, Feb 22, 2012 at 08:55:05AM +0000, Martin J. Evans wrote:
>> On 22/02/12 08:22, Martin J. Evans wrote:
>>> The following rt included a patch to apparently fix that issue - or
>>> perhaps it created a new one, or perhaps it did not get in -
>>> https://rt.cpan.org/Ticket/Display.html?id=74975
>> The patch from that ticket:
>>
>> --- DBI-1.617-Ic0LpE\lib\DBD\File.pm	2010-12-21 23:59:54.000000000 +0100
>> +++ DBI-1.617-n_TYo7\lib\DBD\File.pm	2012-02-13 22:05:24.078125000 +0100
>> -    $str =~ s/([;:])/\\$1/g;
>> +    $str =~ s/([;:\\])/\\$1/g;
>>
>> was applied to subversion trunk by me in revision 15151 14 feb 2012.
>> According to the Changes file 1.616-TRIAL (which I presume is 903 -
>> confusing) was revision 15148 (too early for the fix).
> Ah, yes. I'd forgotten about that patch. Silly of me. I'm sorry for the noise.
>
>>>> The other two reports are 'UNKNOWN' due to compile-time errors:
>>>>
>>>> error C2065: 'my_cxtp' : undeclared identifier
>> I don't see this.
>>
>>>> which seems clearly related to the recent changes.
>>>>
>>>> Given the minor differences in Perl versions (5.12.3 vs 5.12.4) it
>>>> seems like the "OS vers" (5.1 vs 5.2) may be significant here.
>> My Windows machine is v 6.1 according to perl -V and it is Windows 7. My Perl is 5.12.3 and is 64 bit.
>> According to my reading of those UNKNOWN results: 20072290 and 20005629
>> they are both from Leo Susanto
>> one is perl 5.10.1 and the other is 5.12.2
>> both are windows XP (not sure the diff between 5.1 and 5.2)
>> both use the MS Platform SDK for Windows Server 2003 R2 for compiler
>> both use namke
>> both are 32 bit builds
>> both are threaded
> Right. The one that worked is a Win7 Strawberry perl using gcc.
> The failures are using an ActivePerl Build.
>
>> I may be able to find a machine like this.
> That would be great. Thanks.
>
> Hopfully Jan can help since this may be an issue with ActivePerl.
>
> Tim.
>
> p.s. I wish we had more windows users as cpantesters, especially as automated
> smoke testers.
>
I've applied Dave's patch and tested in on Windows with Perl 5.8.9 and 
it now compiles without errors.

Also the patch from Manoj Kumar is applied and this fixes gofer tests.

However, I added these changes to Changes file in the section named

=head2 Changes in DBI 1.618-TRIAL (svn r15148) 13th January 2012

after your (Tim) last release and in reality they were patched after 15148.

I'd suggest you (Tim) did one more development release with the changes 
file amended to move the 2 changes post 15148 into a new versions and I 
will test it on previously failing Windows platforms and whatever other 
platforms I can find. In the mean time I'd also suggest we did not apply 
any more changes.

On the Windows smokers question, I tried that a few years ago and found 
it a PITA eventually bringing my machine to a stand still. To be honest 
I found smoke testing on Linux/AIX/Solaris a PITA too but just not as 
much as Windows. Too many modules prompt even under smoke testing or 
have incorrect dependencies so my exclusion list just kept growing and 
growing. I just looked and I have 32263 smoke reports but you've no idea 
what problems were caused generating them and I eventually gave up - I 
really don't know how bingos (Chris Williams) etc manage to generate so 
many. Personally, I'd change cpan shell to send installation results by 
default but I don't really understand the complexities of this (even if 
they were anonymous).

Martin