Re: hydra and t632

[Efraim Flashner]

On Wed, Oct 21, 2020 at 03:21:55PM -0400, Thien-Thi Nguyen wrote:
> () Efraim Flashner <efraim@flashner.co.il>
> () Wed, 21 Oct 2020 21:38:50 +0300
> 
>    I didn't see any stalling at all so I changed it to 50s and
>    it still finished immediately. I thought there might not be a
>    tty in the build environment but I see in the setup code
>    there is a /dev/tty. It looks like it's bind-mounted from
>    outside the build chroot.
> 
> This tells me that maybe: (a) we are misusing ‘timeout’; or (b)
> ‘timeout’ is not doing its job.  Do any of the following command
> lines result in different behavior?

I inserted time in the build environment for some extra data

>   timeout --foreground 50s        (original)

0.37user 0.14system 0:00.68elapsed 75%CPU (0avgtext+0avgdata 27368maxresident)k
512inputs+1360outputs (0major+17509minor)pagefaults 0swaps
command "time" "make" "check" "KEEPD=1" "VERBOSE=1" "TESTS=t632" failed with 
status 2

>   timeout 50s                     (no --foreground)

0.32user 0.14system 0:00.49elapsed 95%CPU (0avgtext+0avgdata 27372maxresident)k
288inputs+1456outputs (1major+17503minor)pagefaults 0swaps
command "time" "make" "check" "KEEPD=1" "VERBOSE=1" "TESTS=t632" failed with 
status 2

>   LANG=C timeout --foreground 50s (w/ LANG=C)

0.39user 0.12system 0:00.68elapsed 76%CPU (0avgtext+0avgdata 27372maxresident)k
416inputs+1424outputs (2major+17514minor)pagefaults 0swaps
command "time" "make" "check" "KEEPD=1" "VERBOSE=1" "TESTS=t632" failed with 
status 2

>   LANG=C timeout                  (w/ LANG=C, no --foreground)

0.35user 0.10system 0:00.48elapsed 93%CPU (0avgtext+0avgdata 27372maxresident)k
344inputs+1360outputs (2major+16840minor)pagefaults 0swaps
command "time" "make" "check" "KEEPD=1" "VERBOSE=1" "TESTS=t632" failed with 
status 2

And with timeout removed completely:

Command exited with non-zero status 2
0.37user 0.13system 0:00.63elapsed 80%CPU (0avgtext+0avgdata 27372maxresident)k
192inputs+1552outputs (1major+17323minor)pagefaults 0swaps
// output from t632.d/signal-handling-out-and-err:
t632.d/x,v  -->  t632.d/x
revision 1.1 (locked)
writable t632.d/x exists; remove it? [ny](n):
co: checkout aborted

> 
> The reason for ‘LANG=C’ cases is to see if the number parsing is
> sensitive to locale in some weird way (IIRC i saw something
> about that changing in the more recent coreutils NEWS file).
> 
> Another possiblity is (c) ‘co’ is not detecting the TTY, but that
> is doubtful since t632.d/signal-handling-out-and-err has the line
> 
>   writable t632.d/x exists; remove it? [ny](n): 
> 
> which shows ‘co’ initiating user-interaction, which it only does
> if it has detected the TTY.

How does it detect the TTY? It could be that it detects the TTY but it's
not actually useable.

From the build environment:
output of "which tty":
/gnu/store/57xj5gcy1jbl9ai2lnrqnpr0dald9i65-coreutils-8.32/bin/tty

output of "ls -la /dev/tt*":
ls: cannot access '/dev/tt*': No such file or directory

> 
> Hmmm...  i'm hoping for (a), but fear (b) to be the root cause.
> 


-- 
Efraim Flashner   <efraim@flashner.co.il>   אפרים פלשנר
GPG key = A28B F40C 3E55 1372 662D  14F7 41AA E7DC CA3D 8351
Confidentiality cannot be guaranteed on emails sent or received unencrypted

signature.asc
Description: PGP signature