gobject-introspection 1.86.0-1

[Takashi Yano]

On Tue, 17 Mar 2026 16:20:21 +0000
Jon Turney wrote:
> On 17/03/2026 13:04, Takashi Yano via Cygwin wrote:
> > Hi Jon,
> > 
> > Sorry for the late reply.
> 
> No problem! I understand that you have many other, more important 
> demands on your time.
> 
> [...]
> >> Hi Takashi,
> >>
> >> I just tripped over a small problem with gobject-introspection:
> >>
> >>> $ /usr/sbin/alternatives --set python3 /usr/bin/python3.12
> >>>
> >>> $ g-ir-scanner
> >>> Traceback (most recent call last):
> >>>    File "/usr/bin/g-ir-scanner", line 133, in <module>
> >>>      from giscanner.scannermain import scanner_main
> >>>    File "/usr/lib/gobject-introspection/giscanner/scannermain.py", line 35, in <module>
> >>>      from giscanner.ast import Include, Namespace
> >>>    File "/usr/lib/gobject-introspection/giscanner/ast.py", line 27, in <module>
> >>>      from .sourcescanner import CTYPE_TYPEDEF, CSYMBOL_TYPE_TYPEDEF
> >>>    File "/usr/lib/gobject-introspection/giscanner/sourcescanner.py", line 31, in <module>
> >>>      from giscanner._giscanner import SourceScanner as CSourceScanner
> >>> ModuleNotFoundError: No module named 'giscanner._giscanner'
> >>
> >> This fails because the _giscanner module is built for (and linked
> >> against) python3.9.
> > 
> > Thanks for pointing this out.
> > Currently, _giscanner is compiled as
> > /usr/lib/gobject-introspection/giscanner/_giscanner.cpython-39-x86_64-cygwin.dll
> > which perhaps requires libpython3.9.dll.
> > 
> >> I don't quite understand why this is happening since we configure with
> >> "-Dpython=${PYTHON3}", so I'd expect end up with the full path to
> >> specific python version in the shebang line in g-ir-scanner.
> >>
> >> (I think that's what was happening before upstream switched from
> >> autotools to meson)
> >>
> >> In any case, this can be fixed in the cygport by using
> >> python3_fix_shebang on this script.
> > 
> > Could you please let me know bit more detail how to use
> > 'python3_fix_shebang'?
> 
> Sure.
> 
> In this case, I'd suggest adding a custom src_install which contains the 
> default for the meson cygclass [1], then uses python3_fix_shebang[2].
> 
> > src_install() {
> >  cd ${S}
> >  meson_install
> > 
> >  python3_fix_shebang /usr/bin/g-ir-scanner
> > }
> 
> The idea is that this adjusts the shebang line in the installed 
> g-ir-scanner so it references python3.9, so that the import can be found 
> (and thus it can be invoked successfully), irrespective of what version 
> /usr/bin/python3 points to.
> 
> Unfortunately, when I actually try it, it seems that it can't handle 
> '#!/usr/bin/env python3' and just rewrites the env part. How 
> disappointing! That seems like a cygport bug :(
> 
> I guess you could just use something like this instead:
> 
> > sed -i -e "s#python3#${PYTHON3}#" ${D}/usr/bin/g-ir-scanner 

Thanks! I released 1.86.0-2 using:
    sed -i -e "s@python3@${PYTHON3##*/}@" ${D}/usr/bin/g-ir-annotation-tool
    sed -i -e "s@python3@${PYTHON3##*/}@" ${D}/usr/bin/g-ir-doc-tool
    sed -i -e "s@python3@${PYTHON3##*/}@" ${D}/usr/bin/g-ir-scanner

> (But as I say, this all seems a bit like maybe an upstream bug that the 
> python interpreter passed in via CYGMESON_ARGS doesn't end up there anyway?)
> 
> 
> [1] 
> https://cygwin.github.io/cygport/meson_cygclass.html#src_install%20(meson)
> [2] 
> https://cygwin.github.io/cygport/python3_cygclass.html#python3_fix_shebang

Thanks again.

-- 
Takashi Yano <takashi.yano@nifty.ne.jp>