Re: Unbeatable directory leftover from python313 'make test'

[ci4ic4 <ci4ic4%proton.me@localhost>]

On Monday, 6 January 2025 at 21:55, ci4ic4 <ci4ic4%proton.me@localhost> wrote:

> 
>  ... snip 
> > I am about to run the same test under today's -current aarch64. BTW, I ran it earlier under today's amd64 and it completed without any particular issues (well, five tests failed, but I'd say this is a win).
> 
> 
> (Well, ten failed, to be exact)


The mystery deepens.

I am running 'make test' in lang/python313 on two systems, built today from exact same sources - one, amd64, the other, aarch64. Today 'make test' completed on both, with 10 errors on amd64 and 12 - on aarch64.

The issue with the directory with the extended attribute occurs only on the aarch64 system. On the amd64 system after the test execution I can't find any object with schg attribute ( find work | xargs /bin/ls -lo | grep schg ). On the aarch64 system I get four such directories, similar to the one I reported earlier:
...

root ▻ ◇┈◉ find /usr/pkgsrc/lang/python313/work/Python-3.13.1/build/ | xargs /bin/ls -lo | grep schg
drwxr-xr-x  2 root  wheel  schg 512 Jan  6 16:57 @test_13878_tmpæ_SRC
drwxr-xr-x  2 root  wheel  schg 512 Jan  6 22:09 @test_20943_tmpæ_SRC
drwxr-xr-x  2 root  wheel  schg 512 Jan  6 22:23 @test_28978_tmpæ_SRC
drwxr-xr-x  2 root  wheel  schg 512 Jan  6 16:45 @test_3579_tmpæ_SRC 
> 

So I wonder why the test execution here would be so different from under amd64. 


This is the second occasion when I find a difficult to explain difference between the behaviour under these two major architectures - the other being my regular building of lang/zig from the git repo - it stopped working completely somewhere around early October under amd64 and still works just fine under aarch64. I am not even able to bootstrap zig 0.14xxx under NetBSD 10.1 amd64 - which basically means when pkgsrc gets llvm 19, there will be no working zig compiler under amd64, if the things remain as they are (one can't build zig 0.13 with llvm 19). 

Chavdar 


 
> > Thanks again,
> > 
> > Chavdar