mutter 41.2-2 build fail on x86-64-v3 but not on v2 #90

New Issue

Closed

opened 2022-01-08 14:04:14 +01:00 by menter · 6 comments

menter commented

2022-01-08 14:04:14 +01:00

I noticed that mutter was pushed an update, but there was no version for x86-64-v3

I also confirmed that x86-64-v2 was successfully compiled

It looks like the v3 version doesn't pass the final unit tests, so it's considered a failure

I haven't dig into the related question, but I hope this won't be a big deal for us

I noticed that mutter was pushed an update, but there was no version for x86-64-v3 I also confirmed that x86-64-v2 was successfully compiled It looks like the v3 version doesn't pass the final unit tests, so it's considered a failure I haven't dig into the related question, but I hope this won't be a big deal for us

menter changed title from ~~mutter 41 build fail on x86-64-v3 but not on v2~~ to mutter 41.2-2 build fail on x86-64-v3 but not on v2

2022-01-08 14:04:42 +01:00

anonfunc commented

2022-01-08 18:31:23 +01:00

Owner

Let try building it again, see what happens.

menter commented

2022-01-08 18:42:20 +01:00

Author

Let try building it again, see what happens.

I just finished a few compilation tests of v2 and v3...

The first time I compiled with v3 version I reproduced the same result

second time I switched to v2, I also encountered unit tests that failed, but different errors

then I switched to v3 to compile, and I found that it was the same 108th unit test error as the last v2....

after I cleaned the directory again and continued to compile the v3 version, the compilation and testing were completed successfully

I think there must be asynchronous resource management problems that need to be solved

maybe it's possible to repeatedly verify whether there is a probabilistic build failure problem with the x86-64 instruction set, and if the same situation can be reproduced then things will be simpler.

> Let try building it again, see what happens. I just finished a few compilation tests of v2 and v3... The first time I compiled with v3 version I reproduced the same result second time I switched to v2, I also encountered unit tests that failed, but different errors then I switched to v3 to compile, and I found that it was the same 108th unit test error as the last v2.... after I cleaned the directory again and continued to compile the v3 version, the compilation and testing were completed successfully I think there must be asynchronous resource management problems that need to be solved maybe it's possible to repeatedly verify whether there is a probabilistic build failure problem with the x86-64 instruction set, and if the same situation can be reproduced then things will be simpler.

anonfunc commented

2022-01-08 18:48:19 +01:00

Owner

What were your CFLAGS exactly while testing v2/v3?

What were your `CFLAGS` exactly while testing v2/v3?

menter commented

2022-01-08 18:56:36 +01:00

Author

What were your CFLAGS exactly while testing v2/v3?

I simply refer to your settings to modify my environment

CFLAGS="-march=x86-64-v2 -O3 -falign-functions=32 -pipe -fno-plt -fexceptions \
        -Wp,-D_FORTIFY_SOURCE=2 -Wformat -Werror=format-security \
        -fstack-clash-protection -fcf-protection"

and

CFLAGS="-march=x86-64-v3 -O3 -falign-functions=32 -pipe -fno-plt -fexceptions \
        -Wp,-D_FORTIFY_SOURCE=2 -Wformat -Werror=format-security \
        -fstack-clash-protection -fcf-protection"

> What were your `CFLAGS` exactly while testing v2/v3? I simply refer to your settings to modify my environment ``` CFLAGS="-march=x86-64-v2 -O3 -falign-functions=32 -pipe -fno-plt -fexceptions \ -Wp,-D_FORTIFY_SOURCE=2 -Wformat -Werror=format-security \ -fstack-clash-protection -fcf-protection" ``` and ``` CFLAGS="-march=x86-64-v3 -O3 -falign-functions=32 -pipe -fno-plt -fexceptions \ -Wp,-D_FORTIFY_SOURCE=2 -Wformat -Werror=format-security \ -fstack-clash-protection -fcf-protection" ```

anonfunc commented

2022-01-09 14:15:59 +01:00

Owner

There is possibly something time-related going on yes. Maybe its load dependent, not sure. Well, mutterrebuild on x86-64-v3 build fine.

I suggest we keep an eye on this for the next builds.

There is possibly something time-related going on yes. Maybe its load dependent, not sure. Well, `mutter`rebuild on `x86-64-v3` build fine. I suggest we keep an eye on this for the next builds.