gettext, gettext-native-minimal: upgrade 0.23.1 -> 0.26
Quoted 34 lines · Gyorgy Sarvari via lists.openembedded.org
meta/classes-recipe/autotools.bbclass +19
-5@@ -186,13 +186,35 @@ autotools_do_configure() { echo "no" | glib-gettextize --force --copy fi elif [ "${BPN}" != "gettext" ] && grep -q "^[[:space:]]*AM_GNU_GETTEXT" $CONFIGURE_AC; then - # We'd call gettextize here if it wasn't so broken... + # Gettextize could be called here, however it doesn't make the job much easier: + # It doesn't discover relevant po folders on its own, so they still need to be + # found by some heurestics. Also, it would require always the full gettext + # package always, instead of gettext-minimal-native. cp ${STAGING_DATADIR_NATIVE}/gettext/config.rpath ${AUTOTOOLS_AUXDIR}/ - if [ -d ${S}/po/ ]; then - cp -f ${STAGING_DATADIR_NATIVE}/gettext/po/Makefile.in.in ${S}/po/ - if [ ! -e ${S}/po/remove-potcdate.sed ]; then - cp ${STAGING_DATADIR_NATIVE}/gettext/po/remove-potcdate.sed ${S}/po/ + if [ -d ${S}/po ]; then + if [ -f ${S}/po/Makefile.in.in ]; then + # search for all Makefile.in.in files that are identical to ./po/Makefile.in.in, by md5sum + base_makefile_hash=`md5sum ${S}/po/Makefile.in.in | tr -s ' ' | cut -f1 -d' '` + makefiles_to_update=`find ${S} -name Makefile.in.in -exec md5sum {} \+ | grep $base_makefile_hash | tr -s ' ' | cut -d' ' -f2` + else + # if there is no po/Makefile.in.in, then just copy the latest there, without + # searching for others + makefiles_to_update="./po/Makefile.in.in" fi + bbnote List of Makefile.in.ins to update: $makefiles_to_update + for makefile in ${makefiles_to_update}; do + makefile_dir=$(dirname $makefile) + bbnote Executing: cp ${STAGING_DATADIR_NATIVE}/gettext/po/Makefile.in.in ${makefile_dir}/ + cp ${STAGING_DATADIR_NATIVE}/gettext/po/Makefile.in.in ${makefile_dir}/ + if [ ! -e ${makefile_dir}/remove-potcdate.sed ]; then
meta/classes-recipe/autotools.bbclass autotools_do_configure() { · line 219
Quoted 10 lines
meta/classes-recipe/autotools.bbclass +8
-0+ cp ${STAGING_DATADIR_NATIVE}/gettext/po/remove-potcdate.sed ${makefile_dir}/ + fi + done + for makevars in `find ${S} -name Makevars`; do + bbnote Concatenating Makevars: $makevars + cat ${STAGING_DATADIR_NATIVE}/gettext/po/Makevars.template.minimal ${makevars} >> ${makevars}.yocto_temp + mv ${makevars}.yocto_temp ${makevars} + done fi
In principle I think that it's about as I can make it on my own (sans AB failures[1]), but there would be a few question marks on my end:
1. If you have a better idea for relevant Makefile.in.in heurestics in autotools.bbclass, I'm listening
2. I *think* that the Makevars concatenation makes actual sense, and the build success is not only an accident - but I'm not 100% sure if that's the best way to handle this. If you know gettext, I'd like to hear your opinion. (This was inspired by gettextize's behavior. Upon running it, it copies the Makevars.template with default values in the po folder, and asks you to set up all the variables from it in your Makevars file)
[1]: I did build recipes and multiple images successfully, but not "bitbake world" on multiple configs
Quoted 13 lines · Gyorgy Sarvari via lists.openembedded.org
On 8/30/25 22:55, Gyorgy Sarvari via lists.openembedded.org wrote:
In principle I think that it's about as I can make it on my own (sans AB failures[1]), but there would be a few question marks on my end:1. If you have a better idea for relevant Makefile.in.in heurestics in autotools.bbclass, I'm listening
2. I *think* that the Makevars concatenation makes actual sense, and the build success is not only an accident - but I'm not 100% sure if that's the best way to handle this. If you know gettext, I'd like to hear your opinion. (This was inspired by gettextize's behavior. Upon running it, it copies the Makevars.template with default values in the po folder, and asks you to set up all the variables from it in your Makevars file)
Quoted 2 lines
[1]: I did build recipes and multiple images successfully, but not "bitbake world" on multiple configs
Hi Gyorgy,
I took a bit of time to launch a build on the autobuilder. It looks like this is breaking build of several components: glib, e2fsprogs, iso-codes and so on. I saw two types of errors so far:
- undefined symbol: getfilecon
- missing files: t-ca.gmo, t-bg.gmo... I'm not sure where this comes from.
ERROR: glib-2.0-initial-1_2.84.4-r0 do_compile: Execution of '/srv/pokybuild/yocto-worker/no-x11/build/build/tmp/work/x86-64-v3-poky-linux/glib-2.0-initial/2.84.4/temp/run.do_compile.3022462' failed with exit code 127
...
| /srv/pokybuild/yocto-worker/no-x11/build/build/tmp/work/x86-64-v3-poky-linux/glib-2.0-initial/2.84.4/recipe-sysroot-native/usr/bin/msgfmt: symbol lookup error: /srv/pokybuild/yocto-worker/no-x11/build/build/tmp/work/x86-64-v3-poky-linux/glib-2.0-initial/2.84.4/recipe-sysroot-native/usr/bin/../lib/libgettextlib-0.26.so: undefined symbol: getfilecon
ERROR: e2fsprogs-1.47.3-r0 do_compile: Execution of '/srv/pokybuild/yocto-worker/no-x11/build/build/tmp/work/x86-64-v3-poky-linux/e2fsprogs/1.47.3/temp/run.do_compile.3185668' failed with exit code 1
...
| mv: cannot stat 't-ca.gmo': No such file or directory
ERROR: iso-codes-4.18.0-r0 do_compile: Execution of '/srv/pokybuild/yocto-worker/no-x11/build/build/tmp/work/all-poky-linux/iso-codes/4.18.0/temp/run.do_compile.3458421' failed with exit code 1
...
| /srv/pokybuild/yocto-worker/no-x11/build/build/tmp/work/all-poky-linux/iso-codes/4.18.0/recipe-sysroot-native/usr/bin/msgfmt: symbol lookup error: /srv/pokybuild/yocto-worker/no-x11/build/build/tmp/work/all-poky-linux/iso-codes/4.18.0/recipe-sysroot-native/usr/bin/../lib/libgettextlib-0.26.so: undefined symbol: getfilecon
ERROR: Logfile of failure stored in: /srv/pokybuild/yocto-worker/no-x11/build/build/tmp/work/x86-64-v3-poky-linux/psmisc/23.7/temp/log.do_compile.3498395
...
| mv: cannot stat 't-bg.gmo': No such file or directory
You can get all logs here:
https://autobuilder.yoctoproject.org/valkyrie/#/builders/5/builds/2220
https://autobuilder.yoctoproject.org/valkyrie/#/builders/30/builds/2285
https://autobuilder.yoctoproject.org/valkyrie/#/builders/6/builds/2330
https://autobuilder.yoctoproject.org/valkyrie/#/builders/25/builds/2308
We can have more failure logs in all a-full subbuilders, but they all look the same as far as I can see.
https://autobuilder.yoctoproject.org/valkyrie/#/builders/29/builds/2299
Thanks,
Mathieu
Signature · 3 lines
Mathieu Dubois-Briand, Bootlin Embedded Linux and Kernel engineering https://bootlin.com
Quoted 11 lines · Gyorgy Sarvari via lists.openembedded.org
In principle I think that it's about as I can make it on my own (sans AB failures[1]), but there would be a few question marks on my end:
1. If you have a better idea for relevant Makefile.in.in heurestics in autotools.bbclass, I'm listening
2. I *think* that the Makevars concatenation makes actual sense, and the build success is not only an accident - but I'm not 100% sure if that's the best way to handle this. If you know gettext, I'd like to hear your opinion. (This was inspired by gettextize's behavior. Upon running it, it copies the Makevars.template with default values in the po folder, and asks you to set up all the variables from it in your Makevars file)
First, thanks for working on this. Gettext recipe easily wins the 'least sexy' award in oe-core.
Second, the same discussion has stalled in previous update proposal (to 0.24), after Ross suggested we should just try to use gettextize. Ross, can you please comment? We do need to get this update in, and maybe attempts to 'do things right' would be too counter productive.
Alex
Quoted 48 lines · Mathieu Dubois-Briand
Quoted 13 lines · Gyorgy Sarvari via lists.openembedded.org
On 8/30/25 22:55, Gyorgy Sarvari via lists.openembedded.org wrote:
In principle I think that it's about as I can make it on my own (sans AB failures[1]), but there would be a few question marks on my end:1. If you have a better idea for relevant Makefile.in.in heurestics in autotools.bbclass, I'm listening
2. I *think* that the Makevars concatenation makes actual sense, and the build success is not only an accident - but I'm not 100% sure if that's the best way to handle this. If you know gettext, I'd like to hear your opinion. (This was inspired by gettextize's behavior. Upon running it, it copies the Makevars.template with default values in the po folder, and asks you to set up all the variables from it in your Makevars file)Quoted 2 lines
[1]: I did build recipes and multiple images successfully, but not "bitbake world" on multiple configs
Hi Gyorgy,
I took a bit of time to launch a build on the autobuilder. It looks like this is breaking build of several components: glib, e2fsprogs, iso-codes and so on. I saw two types of errors so far:
- undefined symbol: getfilecon
- missing files: t-ca.gmo, t-bg.gmo... I'm not sure where this comes from.ERROR: glib-2.0-initial-1_2.84.4-r0
do_compile: Execution of '/srv/pokybuild/yocto-worker/no-x11/build/build/tmp/work/x86-64-v3-poky-linux/glib-2.0-initial/2.84.4/temp/run.do_compile.3022462' failed with exit code 127
...
|/srv/pokybuild/yocto-worker/no-x11/build/build/tmp/work/x86-64-v3-poky-linux/glib-2.0-initial/2.84.4/recipe-sysroot-native/usr/bin/msgfmt: symbol lookup error:/srv/pokybuild/yocto-worker/no-x11/build/build/tmp/work/x86-64-v3-poky-linux/glib-2.0-initial/2.84.4/recipe-sysroot-native/usr/bin/../lib/libgettextlib-0.26.so: undefined symbol: getfileconERROR: e2fsprogs-1.47.3-r0
do_compile: Execution of '/srv/pokybuild/yocto-worker/no-x11/build/build/tmp/work/x86-64-v3-poky-linux/e2fsprogs/1.47.3/temp/run.do_compile.3185668' failed with exit code 1
...
| mv: cannot stat 't-ca.gmo': No such file or directoryERROR: iso-codes-4.18.0-r0
do_compile: Execution of '/srv/pokybuild/yocto-worker/no-x11/build/build/tmp/work/all-poky-linux/iso-codes/4.18.0/temp/run.do_compile.3458421' failed with exit code 1
...
|/srv/pokybuild/yocto-worker/no-x11/build/build/tmp/work/all-poky-linux/iso-codes/4.18.0/recipe-sysroot-native/usr/bin/msgfmt: symbol lookup error:/srv/pokybuild/yocto-worker/no-x11/build/build/tmp/work/all-poky-linux/iso-codes/4.18.0/recipe-sysroot-native/usr/bin/../lib/libgettextlib-0.26.so: undefined symbol: getfileconERROR: Logfile of failure stored in:
/srv/pokybuild/yocto-worker/no-x11/build/build/tmp/work/x86-64-v3-poky-linux/psmisc/23.7/temp/log.do_compile.3498395
...
| mv: cannot stat 't-bg.gmo': No such file or directoryYou can get all logs here:
https://autobuilder.yoctoproject.org/valkyrie/#/builders/5/builds/2220
https://autobuilder.yoctoproject.org/valkyrie/#/builders/30/builds/2285
https://autobuilder.yoctoproject.org/valkyrie/#/builders/6/builds/2330
https://autobuilder.yoctoproject.org/valkyrie/#/builders/25/builds/2308
Quoted 4 lines
We can have more failure logs in all a-full subbuilders, but they all look the same as far as I can see.
https://autobuilder.yoctoproject.org/valkyrie/#/builders/29/builds/2299
Thanks for taking it for a spin.
This is just a keep-alive ping from my end for now. It is not being ignored, but reproduction is doesn't seem to be straightforward - will report back.
Quoted 3 lines
Thanks,
Mathieu
Quoted 14 lines · Alexander Kanavin
Quoted 11 lines · Gyorgy Sarvari via lists.openembedded.org
In principle I think that it's about as I can make it on my own (sans AB failures[1]), but there would be a few question marks on my end:
1. If you have a better idea for relevant Makefile.in.in heurestics in autotools.bbclass, I'm listening
2. I *think* that the Makevars concatenation makes actual sense, and the build success is not only an accident - but I'm not 100% sure if that's the best way to handle this. If you know gettext, I'd like to hear your opinion. (This was inspired by gettextize's behavior. Upon running it, it copies the Makevars.template with default values in the po folder, and asks you to set up all the variables from it in your Makevars file)
Quoted 7 lines
First, thanks for working on this. Gettext recipe easily wins the 'least sexy' award in oe-core.
Second, the same discussion has stalled in previous update proposal (to 0.24), after Ross suggested we should just try to use gettextize. Ross, can you please comment? We do need to get this update in, and maybe attempts to 'do things right' would be too counter productive.
Yes, briefly I was also looking also into gettextize, and I got to the same conclusion as Robert, it doesn't look that it would have more benefits than drawbacks. The current update logic in autotools.bbclass would be reduced by a line or two, but gettext-minimal-native recipe would be unusable, and all gettext using recipes would have a dependency on the full gettext package.
Nevertheless, if there is interest, I could try to prepare a proposal using gettextize once the current of the patch gets in a state that doesn't blow up on AB.