Yes, it's a bug in libsigc++ 2.3.2. All header files
that contain a specialization of template struct visitor must
include sigc++/visit_each.h, directly or indirectly. This bug is
not noticed if you include sigc++/sigc++.h. I have to check which
header files are affected, and fix this bug.
Thank you for detecting this bug before it got into a stable
version.
Kjell
Den 2014-08-08 01:51,
Ken Moffat skrev:
Hi
I use {,beyond} linuxfromscratch and for most of the packages we
prefer to use stable versions unless there are good reasons. But we
have been using libsigc++-2.3.1 for some time, probably not
realising it was a development version, and we had not noticed any
problems. Until now.
The announcement for 2.3.2 alerted us that this was in fact a
development version, so I was prepared for some testing to see if
things still seemed to work. But instead, I got a build failure in
inkscape (which is my only package that needs the libsigc++...gtkmm
packages).
My initial build used make -j4, and I accidentally overwrote its
log while testing to see if the problem was in gcc-4.9.1,
inkscape-0.48.5, or libsigc++-2.3.2 (all had had point releases
since my previous build in May). I had pasted the final error
message into my notes (lines above that were from other jobs) and
did not realise where the messages had actually started.
I've now taken this to the inkscape list, and provided the log from
a build with make -j1 (160k, so I've gzipped it). One of their devs
says that they do not seem to use visit_each in their code, and he
thinks sigc++ is conflicting with itself whenever a line
#include <sigc++/functors/mem_fun.h>
is reached. He also made a suggestion about changing the inkscape
code to include <sigc++/signal.h> instead, but I figured I would ask
for advice here before perhaps causing breakage for people on an
earlier version of sigc++.
ĸen
|
