Bug#977779: geary FTBFS on mipsel: test suite failure
On Fri, Jan 29, 2021 at 02:11:00PM +0200, Adrian Bunk wrote: > > But this a bug in the CPU, right? Do I understand correctly that the > > package can fail depending on what CPU it is run on regardless of how > > it was built? > > > > I'm trying to understand what we can do in WebKit in order to fix or > > work around this. > > It is a bug in WebKit that a kernel configuration is assumed that is > not even permitted on Loongson: > https://sources.debian.org/src/webkit2gtk/2.31.1-1/Source/WTF/wtf/PageBlock.h/#L54 > https://sources.debian.org/src/linux/5.10.9-1/arch/mips/Kconfig/#L2213-L2215 My question is: I cannot choose a page size depending on the particular MIPS CPU model because once the package is built it could be run on a CPU with a different page size, is that right? If that's the case then I should use a conservative value that is guaranteed to work on all models (64KB, or is 16KB safe?). > BTW: The 4kB setting for arm64 is also questionable, Ubuntu releases > now additionally ship kernels with 64 kB page size: > > https://launchpad.net/ubuntu/groovy/+package/linux-image-unsigned-5.8.0-41-generic-64k Ok, I'll use 64KB then. Berto
Bug#977779: geary FTBFS on mipsel: test suite failure
On Fri, Jan 08, 2021 at 06:07:29PM +0100, Alberto Garcia wrote: > On Mon, Dec 21, 2020 at 11:30:14PM +0200, Adrian Bunk wrote: > > > I see that the build eventually succeeded: > > > > > > > > > https://buildd.debian.org/status/logs.php?pkg=geary=3.38.1-1=mipsel > > > > > > The webkit2gtk build is flaky itself in mipsel, we discussed this > > > already in the past (#962616), I wonder if this is the same root > > > problem? > > > > This is the same problem. > > > > Note that the build is not and never was flaky, it does 100% > > determinisitically fail on Loongson buildds and succeed on Octeon > > buildds. > > > > Jiaxun Yang discovered this weekend that CeilingOnPageSize is wrong > > for Loongson which has 16 kB pagesize and this matches when the > > problems started in 2.28.1, but unfortunately fixing this does not > > seem to fix all problems. > > But this a bug in the CPU, right? Do I understand correctly that the > package can fail depending on what CPU it is run on regardless of how > it was built? > > I'm trying to understand what we can do in WebKit in order to fix or > work around this. It is a bug in WebKit that a kernel configuration is assumed that is not even permitted on Loongson: https://sources.debian.org/src/webkit2gtk/2.31.1-1/Source/WTF/wtf/PageBlock.h/#L54 https://sources.debian.org/src/linux/5.10.9-1/arch/mips/Kconfig/#L2213-L2215 This might not be sufficient to fix all problems of webkit2gtk on MIPS, but has be fixed in WebKit in any case. > Berto cu Adrian BTW: The 4kB setting for arm64 is also questionable, Ubuntu releases now additionally ship kernels with 64 kB page size: https://launchpad.net/ubuntu/groovy/+package/linux-image-unsigned-5.8.0-41-generic-64k
Bug#977779: geary FTBFS on mipsel: test suite failure
On Mon, Dec 21, 2020 at 11:30:14PM +0200, Adrian Bunk wrote: > > I see that the build eventually succeeded: > > > > > > https://buildd.debian.org/status/logs.php?pkg=geary=3.38.1-1=mipsel > > > > The webkit2gtk build is flaky itself in mipsel, we discussed this > > already in the past (#962616), I wonder if this is the same root > > problem? > > This is the same problem. > > Note that the build is not and never was flaky, it does 100% > determinisitically fail on Loongson buildds and succeed on Octeon > buildds. > > Jiaxun Yang discovered this weekend that CeilingOnPageSize is wrong > for Loongson which has 16 kB pagesize and this matches when the > problems started in 2.28.1, but unfortunately fixing this does not > seem to fix all problems. But this a bug in the CPU, right? Do I understand correctly that the package can fail depending on what CPU it is run on regardless of how it was built? I'm trying to understand what we can do in WebKit in order to fix or work around this. Berto
Bug#977779: geary FTBFS on mipsel: test suite failure
On Mon, Dec 21, 2020 at 01:58:32PM +0100, Alberto Garcia wrote: > On Sun, Dec 20, 2020 at 09:09:17PM +0200, Adrian Bunk wrote: > > > Source: geary > > > Version: 3.38.1-1 > > > > The latest version of geary fails to build on mipsel [1]. The test > > > suite fails. > > > > This is not specific to geary and not specific to this version of > > geary, the search is already ongoing for the regression that makes > > webkit2gtk and some rdeps FTBFS on some buildds on mipsel. > > I see that the build eventually succeeded: > > > https://buildd.debian.org/status/logs.php?pkg=geary=3.38.1-1=mipsel > > The webkit2gtk build is flaky itself in mipsel, we discussed this > already in the past (#962616), I wonder if this is the same root > problem? This is the same problem. Note that the build is not and never was flaky, it does 100% determinisitically fail on Loongson buildds and succeed on Octeon buildds. Jiaxun Yang discovered this weekend that CeilingOnPageSize is wrong for Loongson which has 16 kB pagesize and this matches when the problems started in 2.28.1, but unfortunately fixing this does not seem to fix all problems. > Berto cu Adrian
Bug#977779: geary FTBFS on mipsel: test suite failure
On Sun, Dec 20, 2020 at 09:09:17PM +0200, Adrian Bunk wrote: > > Source: geary > > Version: 3.38.1-1 > > The latest version of geary fails to build on mipsel [1]. The test > > suite fails. > > This is not specific to geary and not specific to this version of > geary, the search is already ongoing for the regression that makes > webkit2gtk and some rdeps FTBFS on some buildds on mipsel. I see that the build eventually succeeded: https://buildd.debian.org/status/logs.php?pkg=geary=3.38.1-1=mipsel The webkit2gtk build is flaky itself in mipsel, we discussed this already in the past (#962616), I wonder if this is the same root problem? Berto
Bug#977779: geary FTBFS on mipsel: test suite failure
Control: reassign -1 src:webkit2gtk 2.28.1-1 Control: affects -1 src:geary src:liferea src:evolution src:epiphany-browser On Sun, Dec 20, 2020 at 07:17:13PM +0100, Paul Gevers wrote: > Source: geary > Version: 3.38.1-1 > Severity: serious > Justification: FTBFS > Tags: ftbfs > > Dear maintainer, > > The latest version of geary fails to build on mipsel [1]. The test suite > fails. This is not specific to geary and not specific to this version of geary, the search is already ongoing for the regression that makes webkit2gtk and some rdeps FTBFS on some buildds on mipsel. > Paul cu Adrian
Bug#977779: geary FTBFS on mipsel: test suite failure
Source: geary Version: 3.38.1-1 Severity: serious Justification: FTBFS Tags: ftbfs Dear maintainer, The latest version of geary fails to build on mipsel [1]. The test suite fails. Paul [1] https://buildd.debian.org/status/package.php?p=geary OpenPGP_signature Description: OpenPGP digital signature
