[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Freeze for LLVM packages



Goswin von Brederlow <goswin-v-b@web.de> writes:

> Stéphane Glondu <glondu@debian.org> writes:
>
>> Le 29/08/2010 23:10, Arthur Loiret a écrit :
>>> There you go:
>>> 
>>> http://people.debian.org/~aloiret/squeeze/llvm/llvm-2.6_2.6-10.dsc
>>> http://people.debian.org/~aloiret/squeeze/llvm/llvm-2.6_2.6-10.debdiff
>>
>> From the diff:
>>>  define libllvm-ocaml-dev_extra_binary
>>>  	if test "x$*" = "xlibllvm-ocaml-dev" ; then \
>>> -	  cp $(D)/debian/$*.META $(D)/debian/$*/$(OCAML_STDLIB_DIR)/METAS/META.llvm ; \
>>> +		cp $(D)/debian/$(strip $(call pkgname,$*)).META $(D)/debian/$(strip $(call pkgname,$*))/$(OCAML_STDLIB_DIR)/METAS/META.llvm-$(UVERSION) ; \
>>>  	fi
>>>  endef
>>
>> This won't work because of "." in $(UVERSION). "." is reserved for
>> delimiting subpackages. You should replace them with something else
>> (e.g. "_"), or drop punctuation altogether (as in META.llvm27). And the
>> contents should also reflect the (findlib) package name.
>>
>> Actually, the same applies for llvm-2.7 as well... The META files should
>> be fixed for both, or dropped. They are useless as they are (it was fine
>> when the package name was versionless).
>>
>> Attached is a (tentatively) fixed META.llvm-2_7 file. Beware that the
>> "directory" variable refers to the actual directory (here, llvm-2.7),
>> and the "requires" refer to the package name (i.e. the extension of the
>> META file, here llvm-2_7 or whatever you choose).
>>
>> And by the way, "libllvm-ocaml-2.7-dev" doesn't respect OCaml packaging
>> policy (which is still work in progress...) and is not handled by
>> dh_ocaml (the virtual package with ABI hash is not provided)...
>> something like "libllvm-2.7-ocaml-dev" would be better.
>>
>>
>> Cheers,
>>
>> -- 
>> Stéphane
>
> Should we have an ExtUnix.Linux submodule for eventfd and signalfd?
>
> Or use
>
> let eventfd () = throw ENOSYS
>
> on platforms that don't have it?
>
>
> I could imagine emulating the eventfd with pipes but signalfd might be
> impossible to emulate.
>
> MfG
>         Goswin

Please ignore. I somehow replied to the wrong mail.

MfG
        Goswin


Reply to: