Re: enerating Shell-0.1.gir with a custom command. FAILED: src/Shell-0.1.gir



On Mon, 2017-09-25 at 19:03 +0200, tjoen dds nl wrote:
Giovanni Campagna schreef op 25-09-2017 17:57:
On Mon, 2017-09-25 at 12:42 +0200, tjoen dds nl wrote:
System: Linux LFS-like
gnome-shell-3.26.0
gobject-introspection-1.54.0 make check passed all tests
bash-4.4 with all 12 patches

cut and paste from xterm

[...]
Command
'[u'/home/tdance/rpmbuild/BUILD/gnome-shell-3.26.0/_build/tmp-
introspectJOKYck/Shell-0.1',
u'--introspect-dump=/home/tdance/rpmbuild/BUILD/gnome-shell-
3.26.0/_build/tmp-
introspectJOKYck/functions.txt,/home/tdance/rpmbuild/BUILD/gnome-
shell-3.26.0/_build/tmp-introspectJOKYck/dump.xml']'
returned non-zero exit status -11
ninja: build stopped: subcommand failed.

Similar problem with gjs-1.50.0 : Command
'[u'/home/tdance/rpmbuild/BUILD....
where I circumvented problem with
$ sed -i 's/INTROSPECTION_GIRS = GjsPrivate-
1.0.gir/INTROSPECTION_GIRS
=/g' Makefile.in

I am not familiar enough with meson and ninja to find a
circumvention
or
how to disable GIR

You cannot disable GIR when you build gnome-shell, because gnome-
shell
is partially written in JavaScript and needs GIR information to
load
its C library.

Ai! that means I am stuck here. I wonder if it is possible using 
gnome-shell,
and gnome-session from Gnome-3.24.2

All versions of gnome-shell (dating back to 3.0) need gobject-
introspection to work.

From the error code, -11, it might be that some process is
segfaulting
in the introspection process. Specifically, it looks like the
temporary
binary that is run to extract property information is segfaulting.
A reason for this could be lacking a working display server on the
build machine (some gtk libraries don't like having their classes
probed if gtk is not initialized), but I'm not sure.

It compiled under startx with mutter as window manager.

Investigating further what is segfaulting might help you.

In build.ninja:
rule CUSTOM_COMMAND
  command = $COMMAND

The command i '[u'...
I have never seen such command. Is that command normal?

Sort of.

ninja just invokes g-ir-scanner. g-ir-scanner in turns compiles a small
program in a temporary directory and then invokes it.
It's this last invocation in the temporary directory that's failing.

The command looks weird because it's already split into multiple
arguments (hence the []) and it's using python2 and unicode objects
(hence the u'' syntax).

Giovanni

Attachment: signature.asc
Description: This is a digitally signed message part



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