[opensuse-factory] Re: Dependencies of python3-qt5-devel, hplip build failures for ppc64 and armv7l
On 8/27/19 8:51 PM, Martin Wilck wrote:
Hi,
the hplip build fails on openSUSE:Factory:PowerPC because of the following dependency change:
hplip (buildreq) -> python3-qt5-devel -> libqt5-qttools-devel -> libqt5-qttools-doc -> clang = 8.0.0
... because clang is unavailable on ppc64.
It looks like qdoc support is the only thing that requires clang, updating the bcond at the top of https://build.opensuse.org/package/view_file/openSUSE:Factory/libqt5-qttools... to exclude building qdoc support on ppc64 looks like it should fix this issue.
Likewise, the following chain breaks the build on armv7l:
hplip (buildreq) -> python3-qt5-devel -> libqt5-qtwebengine-devel -> libavcodec58 -> libdav1d.so.1
It seems that I can fix my build using #!BuildIgnore. Yet I'd like to question whether these are really hard dependencies. Is it reasonable to require libavcodec to be installed in order to compile for Qt5?
qtwebengine embeds a modified chromium browser into applications, chromium seems to use libavcodec rather then gst for audio so libavcodec ends up as a runtime dependency of qtwebengine as well. I imagine there are more applications that use qtwebengine that require working audio then there are python3-qt5 applications that are using qtwebengine, so maybe it would make sense to split qtwebengine out of python3-qt5 into a sub package. At the same time we don't build qtwebengine for a number of architectures such as ppc64 which is maybe why you have less issues there if aarch64 still works given many people use it on the Raspberry Pi, maybe we should look at dropping qtwebengine for armv7l if its not going to work (i'd take a guess people probably haven't tested it there). Cheers Simon -- Simon Lees (Simotek) http://simotek.net Emergency Update Team keybase.io/simotek SUSE Linux Adelaide Australia, UTC+10:30 GPG Fingerprint: 5B87 DB9D 88DC F606 E489 CEC5 0922 C246 02F0 014B -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org
Hi! On 8/27/19 2:26 PM, Simon Lees wrote:
... because clang is unavailable on ppc64.
LLVM and Clang are fully supported by upstream on ppc64 and ppc32 (both big-endian). If it's not building on openSUSE, then you've found a bug that should be fixed.
It looks like qdoc support is the only thing that requires clang, updating the bcond at the top of https://build.opensuse.org/package/view_file/openSUSE:Factory/libqt5-qttools... to exclude building qdoc support on ppc64 looks like it should fix this issue.
No, you should fix LLVM/clang on ppc64 if it's broken there.
Likewise, the following chain breaks the build on armv7l:
hplip (buildreq) -> python3-qt5-devel -> libqt5-qtwebengine-devel -> libavcodec58 -> libdav1d.so.1
It seems that I can fix my build using #!BuildIgnore. Yet I'd like to question whether these are really hard dependencies. Is it reasonable to require libavcodec to be installed in order to compile for Qt5?
qtwebengine embeds a modified chromium browser into applications, chromium seems to use libavcodec rather then gst for audio so libavcodec ends up as a runtime dependency of qtwebengine as well.
And I'm very confident that both ffmpeg and qtwebengine work fine on ARMv7 as they do on Debian.
I imagine there are more applications that use qtwebengine that require working audio then there are python3-qt5 applications that are using qtwebengine, so maybe it would make sense to split qtwebengine out of python3-qt5 into a sub package. At the same time we don't build qtwebengine for a number of architectures such as ppc64 which is maybe why you have less issues there if aarch64 still works given many people use it on the Raspberry Pi, maybe we should look at dropping qtwebengine for armv7l if its not going to work (i'd take a guess people probably haven't tested it there).
You can see in Debian on which architecture qtwebengine is supported:
https://buildd.debian.org/status/package.php?p=qtwebengine-opensource-src&suite=experimental
Google is still maintaining ARMv7 support in the foreseeable future, so it should not be disabled there. Adrian N�����r��y隊Z)z{.���r�+�맲��r��z�^�ˬz��N�(�֜��^� ޭ隊Z)z{.���r�+��0�����Ǩ�
On Tue, 2019-08-27 at 12:37 +0000, John Paul Adrian Glaubitz wrote:
No, you should fix LLVM/clang on ppc64 if it's broken there.
Thanks for the hint. At a second look, clang8 does indeed exist for ppc64. But it's clang8 = 8.0.1, and libqt5-qttools-doc requires clang8 = 8.0.0, because of this code in the spec file: # qdoc hardcodes clang include paths: boo#1109367, QTBUG-70687 %requires_eq clang%(rpm -q --qf '%''{version}' clang-devel | cut -d. -f1) libqt5-qttools apparently hasn't been rebuild with clang8-8.0.1 because of a build failure on July 30th (!). The last successful build was on June 28th. At first sight, the current build failure seems to be LTO related ("undefined reference to `LTHUNK9.lto_priv.2'"). Currently the package is blocked, one needs to have patience for ppc64. I've opened bug 1148374. Regards Martin
participants (3)
-
John Paul Adrian Glaubitz -
Martin Wilck -
Simon Lees