Hi Tobias,
> Small re-diff - but essentially unchanged.
>
> (I made a thinko when adding a _final call to
> gfc_trans_class_array_init_assign: Not in all contexts the _final should be
> called, only for INTENT(OUT). Thus, I remove the _final call and deferred it
> to the actual finalization call. [That also matches the scalar handling,
> which only does a memcpy and no dealloc.])
>
>
> Build and regtested on x86-64-gnu-linux.
> OK for the trunk?
I think this patch is ok. Just one nit:
@@ -5571,7 +5569,7 @@ gfc_dump_module (const char *name, int dump_flag)
FIXME: For backwards compatibility with the old uncompressed
module format, write an extra empty line. When the module version
is bumped, this can be removed. */
- gzprintf (module_fp, "GFORTRAN module version '%s' created from %s\n\n",
+ gzprintf (module_fp, "GFORTRAN module version '%s' created from %s\n",
MOD_VERSION, gfc_source_file);
Here you should remove the FIXME.
Thanks for the patch,
Janus
> Tobias Burnus wrote:
>>
>> Pre-remark: This patch does *not* enable finalization or polymorphic
>> deallocation.
>>
>> * * *
>>
>> Dear all,
>>
>> The attached patch is a bit boring and invasive, but it paves the way to
>> FINAL support.
>>
>> Changes of technical kind:
>>
>> * Changed ABI for CLASS's virtual table (due to _final) - and, hence, it
>> bumps the .mod version
>> * The finalization wrapper is now generated (this should not but might
>> lead to ICEs)
>> * It also causes that the virtual table is now more often generated
>>
>> New feature:
>>
>> _copy no longer deallocates the "dst" argument. Doing so lead to bogus
>> finalization with ALLOCATE (exposed with the pending FINAL patch). As a
>> sideeffect, memset could be removed and CALLOC could be replased by MALLOC
>> (minute performance advantage). In order to keep the deallocation in
>> gfc_trans_class_array_init_assign, there is now a call to the finalization
>> wrapper.
>>
>> Next steps:
>> * Add end-of-scope/intent(out) deallocation for polymorphic arrays
>> * Enable FINAL parsing
>> * Stepwise enabling for polymorphic deallocation/finalization
>> * Fix issues with ELEMENTAL(+optional) with intent(out)
>> * Fix some issues related to intrinsic assignment
>> * Fix fallout of any of those items
>>
>> Build and regtested on x86-64-gnu-linux.
>> OK for the trunk?
>>
>> Tobias
