Re: CVS commit: pkgsrc/print

[Frédéric Fauberteau <triaxx%NetBSD.org@localhost>]

Le 26/05/2020 à 21:37, Leonardo Taccari a écrit :
> Hello Frédéric, 
> 
> Frédéric Fauberteau writes:
>> [...]
>> Log Message:
>> libcups: fix build failure on FreeBSD
>>
>> pkgsrc changes:
>> ---------------
>>   * Add conditional inclusion for FreeBSD.
>>   * Apply some changes to appease pkglint.
>>   * Bump revision.
>> [...]
>> @@ -1,4 +1,4 @@
>> -# $NetBSD: buildlink3.mk,v 1.15 2020/05/22 10:55:49 adam Exp $
>> +# $NetBSD: buildlink3.mk,v 1.16 2020/05/26 08:36:55 triaxx Exp $
>>  
>>  BUILDLINK_TREE+=    cups-base
>>  
>> @@ -6,21 +6,21 @@ BUILDLINK_TREE+=   cups-base
>>  CUPS_BASE_BUILDLINK3_MK:=
>>  
>>  BUILDLINK_API_DEPENDS.cups-base+=   cups-base>=1.1.19nb3
>> -BUILDLINK_ABI_DEPENDS.cups-base+=   cups-base>=2.3.3nb3
>> +BUILDLINK_ABI_DEPENDS.cups-base+=   cups-base>=2.3.3nb4
>>  BUILDLINK_PKGSRCDIR.cups-base?=             ../../print/cups-base
> 
> Why BUILDLINK_ABI_DEPENDS.cups-base was bumped? (if the commit just
> fixed support on FreeBSD and possibly ends up in binary change for
> cups-base package on other platforms, only a PKGREVISION bump is
> needed)

After reading you, I am convinced that bumping of BUILDLINK_ABI_DEPENDS.cups-base was probably not needed. But when I commit, the fact that BUILDLINK_ABI_DEPENDS.cups-base was the same as PKGREVISION 
let me think that it would be matching. In addition, I read too fast the passage containing "[...] need their PKGREVISIONs increased and, if they have buildlink3.mk files, their 
BUILDLINK_ABI_DEPENDS.pkg adjusted, too."

It reminds me that I have always found that manipulating BUILDLINK_{API,ABI}_DEPENDS was a little bit touchy.
 
>> [...]
>> -.if ${OPSYS} != "Darwin" && !empty(PKG_BUILD_OPTIONS.cups-base:Mavahi)
>> +.if ${OPSYS} != "Darwin" && ${PKG_BUILD_OPTIONS.cups-base:Mdnssd}
>>  .include "../../net/avahi/buildlink3.mk"
>>  .endif
>> [...]
> 
> Is this a copypasto?  I.e. it probably needs an s/dnssd/avahi/? (or why
> was it changed?)

It's clearly a copypasto that went wrong. Thank you for your alertness.
 
>> @@ -6,21 +6,21 @@ BUILDLINK_TREE+=   libcups
>>  LIBCUPS_BUILDLINK3_MK:=
>>  
>>  BUILDLINK_API_DEPENDS.libcups+=     libcups>=1.1.19nb3
>> -BUILDLINK_ABI_DEPENDS.libcups+=     libcups>=2.3.3nb2
>> +BUILDLINK_ABI_DEPENDS.libcups+=     libcups>=2.3.3nb3
>>  BUILDLINK_PKGSRCDIR.libcups?=       ../../print/libcups
>> [...] 
> 
> Same comments regarding cups-base, probably only PKGREVISION bump was
> needed not BUILDLINK_ABI_DEPENDS.libcups bump.
> 
> 
> Regarding the BUILDLINK_ABI_DEPENDS.{cups-base,libcups} bump too much
> time has passed so it could not be downgraded as it would be more
> problematic, but regarding the `s/avahi/dnssd/' change in
> print/cups-base/buildlink3.mk' logic can you please elaborate further
> or revert it to check the `avahi' option as it was?
> 
> 
> Thanks!
>