On 09/02/2012 22:12, Tim Bunce wrote:
> On Thu, Feb 09, 2012 at 09:53:08PM +0000, Martin J. Evans wrote:
>> Something between 1.616_901 and 1.616_902 changed which breaks fork() in DBI on Windows. The test 16destroy.t fails with a Windows error
>>
>> "This application has requested the Runtime to terminate in an unusual way"
> I'd guess that's Windows polite version of Seg Fault ot Bus Error :)
> I know ~0 about windows but I seem to recall that there _might_ be
> something useful in the system event logs.
I'll look more but no luck so far.
>> Initially I saw this in 1.617 but then I stepped backwards to find the problem occurred between these 2 versions. Some smokers have reported this problem in 5.10.1, 5.14.0 and 5.14.2. I can reproduce on my wife's netbook which is strawberry perl 5.12.0.
>>
>> As you'll see from the Changes file the only 2 entries added were:
>>
>> Enhanced performance for threaded perls (Dave Mitchell, Tim Bunce)
>> Added note to DBI::Profile about async queries (Marcel Gr<FC>nauer).
>>
>> I've diffed the two versions but there were quite a lot of changes and
>> so far I cannot locate the problem. I've given up for tonight but
>> might find time to do more.
> When you get a moment could you try this change:
>
> --- DBI.xs (revision 15137)
> +++ DBI.xs (working copy)
> @@ -3536,7 +3536,7 @@
> }
> }
>
> - if (is_orig_method_name)
> + if (0&& is_orig_method_name)
> imp_msv = (SV*)inner_method_lookup(aTHX_ DBIc_IMP_STASH(imp_xxh),
> cv, meth_name);
> else
>
> which will, I think, effectively disable the runtime side of the method
> cache changes.
>
> Tim.
nope, didn't change anything - it still failed.
Martin
|