Hello,
<jann...@gnu.org> skribis:
>>> Anyway, using this patch 0001 it seems that suppressing the warnings
>>> works, I no longer get
>>>
>>> "GC Warning: Repeated allocation of very large block (appr. size 112
>>> KiB):\n\tMay lead to memory leak and poor performance\n"
>>>
>>>
>>> but still get
>>>
>>> unexpected build daemon error: stoi
>>
>> Damnit. Could you check with rpctrace what the daemon receives?
>>
>> I wonder if I misunderstood what the root cause is.
>
> Yes :-( I captured a `guix offload test' run, see attached.
[...]
> task31(pid198)-> 2058 (4) = 0 22<--9(pid198)
> 22<--9(pid198)->exec_startup_get_info () = 0 134517280 134512692 352 237568
> 16777216 0
> "/gnu/store/7wgwfsbvq8m9zkz03d27ij53jciliz9n-guix-1.4.0-27.3d399e5/libexec/guix/guile\0\\0/run/current-system/profile/bin/guix\0authenticate\0"
>
> "SHELL=/gnu/store/dm5shwb20i38wqdkmyqvhqfi0hmq1lr1-bash-5.1.16/bin/bash\0XDG_CONFIG_DIRS=/root/.guix-profile/etc/xdg:/run/current-system/profile/etc/xdg\0PKG_CONFIG_PATH=/run/current-system/profile/lib/p"
> { 16<--25(pid198) 13<--27(pid198) 13<--27(pid198) 4<--32(pid198)
> 12<--33(pid198)} { 11<--34(pid198) 6<--35(pid198) 2<--36(pid198)
> 26<--38(pid198) (null) (null)} {18 0 0 0 0}
[...]
> 26<--38(pid198)->proc_setmsgport_request ( 44<--48(pid-1)) = 0 (null)
> 26<--38(pid198)->proc_set_arg_locations_request (17014228 17014248) = 0
> task31(pid198)-> 3204 (1) = 0 pn{ 19}
> task31(pid198)-> 3215 (pn{ 19} 49) = 0
> task31(pid198)-> 3204 (1) = 0 pn{ 20}
> task31(pid198)-> 3210 (pn{ 20} 1) = 0
> 26<--38(pid198)->proc_handle_exceptions_request ( 49<--51(pid-1)
> 50<--52(pid-1) 5 {75 31 31 31 0 0 0 0 0 0 0 0 19291280 23 0 21548192 0}) = 0
> thread46(pid198)-> 2068 (3 51) = 0
> task31(pid198)-> 3206 (pn{ 19}) = 0
> task31(pid198)-> 2023 (17031168 20) = 0
> task31(pid198)-> 2023 (17027072 24) = 0
> 6<--35(pid198)->dir_lookup
> ("gnu/store/81ffz0prarfczr408ydnps31jf72s5ly-glibc-cross-i586-pc-gnu-2.39/share/locale/locale.alias"
> 4194305 0) = 0 1 "" 51<--47(pid198)
Does that ‘locale.alias’ file exists?
Did you try several LC_ALL=xxx values to see which one would work and
which one wouldn’t?
So after all, there may be two issues: the “Repeated allocation” thing,
and a locale issue.
Thanks,
Ludo’.
