Building minicom package fails when the output directory path contains the string "m4". e.g. make O=output-m4 qemu_x86_64_defconfig make minicom The build output looks like: >>> minicom a6a507dcb8bd4cb5da7af9e2dc75b8e495fedbb6 Configuring >>> minicom a6a507dcb8bd4cb5da7af9e2dc75b8e495fedbb6 Autoreconfiguring configure.ac:27: warning: macro 'AM_ICONV_LINK' not found in library configure.ac:125: warning: macro 'AM_GNU_GETTEXT' not found in library configure.ac:126: warning: macro 'AM_GNU_GETTEXT_VERSION' not found in library autoreconf: configure.ac: AM_GNU_GETTEXT_VERSION is used, but not AM_GNU_GETTEXT configure.ac:27: error: possibly undefined macro: AM_ICONV_LINK If this token and others are legitimate, please use m4_pattern_allow. See the Autoconf documentation. configure.ac:125: error: possibly undefined macro: AM_GNU_GETTEXT configure.ac:126: error: possibly undefined macro: AM_GNU_GETTEXT_VERSION autoreconf: /home/hthiery/sources/mainline/buildroot/output-m4/host/bin/autoconf failed with exit status: 1 make[1]: *** [package/pkg-generic.mk:249: /home/hthiery/sources/mainline/buildroot/output-m4/build/minicom-a6a507dcb8bd4cb5da7af9e2dc75b8e495fedbb6/.stamp_configured] Fehler 1 make: *** [Makefile:23: _all] Fehler 2 I also observed that a directory is created where the "m4" part of the path is replaced by the next part of the host directory. This directory contains links to m4 files. In this example: output-host/ output-host/share output-host/share/aclocal output-host/share/aclocal/codeset.m4 output-host/share/aclocal/wchar_t.m4 output-host/share/aclocal/fcntl-o.m4 output-host/share/aclocal/size_max.m4 output-host/share/aclocal/lib-link.m4 output-host/share/aclocal/progtest.m4 output-host/share/aclocal/intdiv0.m4 output-host/share/aclocal/iconv.m4 output-host/share/aclocal/uintmax_t.m4 output-host/share/aclocal/xsize.m4 output-host/share/aclocal/inttypes-pri.m4 output-host/share/aclocal/intldir.m4 output-host/share/aclocal/lock.m4 output-host/share/aclocal/longlong.m4 output-host/share/aclocal/nls.m4 output-host/share/aclocal/glibc21.m4 output-host/share/aclocal/lib-prefix.m4 output-host/share/aclocal/visibility.m4 output-host/share/aclocal/lib-ld.m4 output-host/share/aclocal/configheader.m4 output-host/share/aclocal/wint_t.m4 output-host/share/aclocal/printf-posix.m4 output-host/share/aclocal/intlmacosx.m4 output-host/share/aclocal/threadlib.m4 output-host/share/aclocal/stdint_h.m4 output-host/share/aclocal/lcmessage.m4 output-host/share/aclocal/intmax.m4 output-host/share/aclocal/po.m4 output-host/share/aclocal/ansi-c++.m4 output-host/share/aclocal/intl.m4 output-host/share/aclocal/glibc2.m4 output-host/share/aclocal/inttypes_h.m4 output-host/share/aclocal/gettext.m4 The same can be reproduced by: utils/test-pkg -d ~/pkg-test-m4 -p minicom
The issues is also seens with the "pdbg" and "dropwatch" packages.
Hi Heiko, I've just tried to rebuild with gettext instead of gettext-tiny, so at the moment it seems a gettext-tiny problem. I'm going to dig deeper and in the meanwhile I opened an issue on gettext-tiny Github: https://github.com/sabotage-linux/gettext-tiny/issues/48
It seems to be caused by host-gettext-tiny, which installs stuff in the wrong directory, so the relevant m4 files can't be found. I've done a quick check and found no upstream commits or PRs or issues that seem relevant. This one looks funny: gettext-gnu/gettext-tools/misc/autopoint.in: m4/ ) destfile=`echo "$1" | sed -e "s,^m4/,$m4dir/,"` ;;
Heiko, you've discovered a bug that is covered at the moment, since minicom needs NLS, then gettext(entire). So at the moment the bug is hidden. Arnout, thank you for pointing that. Maybe I'm going to manage it and open PR ;-)
This happens only with folder ending with "m4". If you try rebuild with: make O=output-m4-tiny qemu_x86_64_defconfig make minicom It passes that point.
(In reply to Giulio Benetti from comment #5) I see .. thank you.
I've opened a PR to fix it: https://github.com/sabotage-linux/gettext-tiny/pull/49 This solves the bug. I'm going to send the patch to Buildroot ML. Heiko, can you please check if this patch solve the problem? Thank you
I've wrongly marked as fixed. This is the patch for Buildroot to test: https://patchwork.ozlabs.org/patch/1224921/ Best regards
Fixed with: https://git.buildroot.org/buildroot/commit/?id=41b9a64526b7bbd28ba2d084f3786cb536750daa Thanks both for the report, investigation and fix! :-)