severity 362657 normal
tags 362657 + moreinfo
stop
Hi,
I'm trying to respond to bugs you filed against libmms individually,
here I'm only responding to concerns you expressed in #362657.
On Fri, Apr 14, 2006, Wesley J. Landaker wrote:
> The public headers for libmms in libmms-dev include "bswap.h", which
> includes "glib.h". This is a needless dependency, because nothing in the
> *headers* actually uses any of the macros from bswap, nor do they use
> anything from glib.h.
libmms-dev ships bswap.h because it is included from the API headers of
libmms. This makes anything in bswap.h part of the libmms API. In
particular, this means the LE_16, BE_16, LE_32, BE_32, LE_64, and BE_64
are part of the API of libmms; I'm afraid this also pulls in the whole
glib API, since it's #included, but this is subject to debates.
> I realize that in *building* libmms, it uses some macros from glib, but
> since those macros are not needed or referenced by any users of libmms,
> this needs to be fixed.
It's too late to remove them from the public API that the libmms
developers have chosen to expose. All I can propose concerning this
particular point is to suggest the upstream developers to stop defining
the macros in the libmms API for the next major version of libmms that
will break API.
> Alternately, libmms-dev could have a real package dependency on the
> correct glib development package, but since libmms itself doesn't link
> to glib, this would be very silly. =)
I agree that libmms-dev lacks a libglib2.0-dev dependency, especially
since the glib headers are sourced from the libmms headers, and since
the glib pkg-config file is required by the libmms pkg-config file.
I'll do an upload shortly to address this issue.
Do you agree that #362657 is only about telling upstream they should
avoid exposing the macros I quoted and the glib API when it seems it's
not truly required?
Bye,
--
Loïc Minier <[EMAIL PROTECTED]>
"You can gtk_main_run, but you can't gtk_widget_hide." --danw, 19-jul-04
