Re: CVS commit: pkgsrc/mail/fdm

[Joerg Sonnenberger <joerg%britannica.bec.de@localhost>]

On Thu, Oct 10, 2013 at 10:05:01AM +0900, OBATA Akio wrote:
> On Thu, 10 Oct 2013 09:38:54 +0900, Joerg Sonnenberger 
> <joerg%britannica.bec.de@localhost> wrote:
> 
> >On Thu, Oct 10, 2013 at 09:37:12AM +0900, OBATA Akio wrote:
> >>On Thu, 10 Oct 2013 08:22:13 +0900, Joerg Sonnenberger 
> >><joerg%britannica.bec.de@localhost> wrote:
> >>
> >>>On Thu, Oct 03, 2013 at 11:12:43AM +0000, OBATA Akio wrote:
> >>>>Module Name:      pkgsrc
> >>>>Committed By:     obache
> >>>>Date:             Thu Oct  3 11:12:43 UTC 2013
> >>>>
> >>>>Modified Files:
> >>>>  pkgsrc/mail/fdm: Makefile PLIST distinfo
> >>>>  pkgsrc/mail/fdm/patches: patch-aa patch-fdm.h
> >>>>Added Files:
> >>>>  pkgsrc/mail/fdm: options.mk
> >>>>
> >>>>Log Message:
> >>>>Update fdm to 1.7.
> >>>>Based on PR pkg/48254 by Leonardo Taccari.
> >>>
> >>>clang -I. -I- -I/usr/local/include -I/usr/pkg/include -I/usr/include 
> >>>-I/usr/pkg/include -O2 -I/usr/pkg/include -I/usr/include -DBUILD="\"1.7\"" 
> >>>-DPCRE -c child-deliver.c -o child-deliver.o
> >>>clang: error: '-I-' not supported, please use -iquote instead
> >>>*** [child-deliver.o] Error code 1
> >>
> >>Isn't it a job for wrapper?
> >
> >No. I'd go the step further and say that something is seriously broken
> >if it needs -I- or -iquote in first place.
> 
> He? clang itself claims to use `-iquote' instead of '-I-', so such 
> transformation
> should be done in mk/wrapper/transform-clang?

Not easily. The problem with -I- is that it changes the semantics of the
other -I depending on the place it is used. That's highly nasty. Most
other compilers don't even support either form.

> And again and again, if you think it is broken, YOU MUST REPORT TO UPSTREAM!
> Such matter is not a packaging issue and pkgsrc is not only for clang.

I don't have time to deal with hundreds of different upstream vendors.
Whether they care about anything beyond Linux/GCC is impossible to tell
in advance and in that sense it *is* a packaging issue. As I said, this
is not just a problem with clang. It is bad practise in general. Have
you even checked what -I- and/or -iquote do before saying that it is a
clang problem?

Joerg