bug#36471: Unreadable Stacktrace Example

[Ludovic Courtès]

Hi Björn,

I’m reassigning this bug report to ‘guile’.  :-)

Björn Höfling <address@hidden> skribis:

> First the full stacktrace, below are some comments:
>
> ./pre-inst-env  guix lint -c cve freecad                         
> Backtrace:reecad@0.18.2-1.dbb4cc6 [cve]...
>           11 (apply-smob/1 #<catch-closure 1a071a0>)
> In ice-9/boot-9.scm:
>     705:2 10 (call-with-prompt _ _ #<procedure default-prompt-handle?>)
> In ice-9/eval.scm:
>     619:8  9 (_ #(#(#<directory (guile-user) 1ac5140>)))
> In guix/ui.scm:
>   1668:12  8 (run-guix-command _ . _)
> In srfi/srfi-1.scm:
>     640:9  7 (for-each #<procedure 1df01a0 at guix/scripts/lint.scm?> ?)
> In guix/scripts/lint.scm:
>    1152:4  6 (run-checkers #<package freecad@0.18.2-1.dbb4cc6 gnu/p?> ?)
> In srfi/srfi-1.scm:
>     640:9  5 (for-each #<procedure 3e473c0 at guix/scripts/lint.scm?> ?)
> In guix/scripts/lint.scm:
>     933:4  4 (check-vulnerabilities _)
>     928:9  3 (_ _)
> In unknown file:
>            2 (force #<promise #<procedure 7f1ce3ab6228 at guix/scrip?>)
> In guix/scripts/lint.scm:
>    917:24  1 (_)
> In ice-9/boot-9.scm:
>     829:9  0 (catch srfi-34 #<procedure 37e64e0 at guix/scripts/lin?> ?)
>
> ice-9/boot-9.scm:829:9: In procedure catch:
> In procedure connect: Connection timed out

[...]

> In guix/scripts/lint.scm:
>     933:4  4 (check-vulnerabilities _)
>     928:9  3 (_ _)
>
> --> 933: Nice, I'm in check-vulnerabilities. But why is there no argument?
> --> What is this strange second line: (_ _)? Why is there nothing written?

Good question, I’m not sure why that happens, but hopefully it’s easily
fixed.

> In unknown file:
>            2 (force #<promise #<procedure 7f1ce3ab6228 at guix/scrip?>)
>
> --> Why is this in a "unknown file"?

Because ‘force’ is implemented in C, so there’s no source location info
available.

> --> Is this force/promise making my stacktrace more unreadable?

No.

> In guix/scripts/lint.scm:
>    917:24  1 (_)
>
> --> Uh. Any more details?. Is "(_)" the call to "(current-vulnerabilities*)"?!

Dunno.  :-/

> In ice-9/boot-9.scm:
>     829:9  0 (catch srfi-34 #<procedure 37e64e0 at guix/scripts/lin?> ?)
>
> ice-9/boot-9.scm:829:9: In procedure catch:
> In procedure connect: Connection timed out
>
> --> Now, wait: The real problem happens somewhere down here and I don't
> have any detailed stacktrace about that?
>
> I have to manually go down into current-vulnerabilities*,
> call/nw/failsaife, etc? Why don't I get the details from here?

Part of what makes information “disappear” from stack traces is the
tail-call optimization (TCO).  If F calls G, and G then calls H in tail
position, then the backtrace will suggest that F called H (G doesn’t
appear at all.)

IIRC Chez Scheme has a fancy trick that allows it to keep track of part
of the “conceptual” stack trace, without losing the space guarantees
that TCO provides.  It cannot come for free though, but I wonder if
Guile should provide it in ‘--debug’ mode or something.

> Is this because in "call-with-networking-fail-safe", line 900 we
> through newly:
>
>         (args
>          (apply throw args))))))
>
> but then the stacktrace gets lost?
>
> Would there be a throw-with-caused-by?

Not sure what you mean, but note that in many cases ‘throw’ itself is
called in tail position, which can also contribute to obfuscating the
stack trace…

Thanks,
Ludo’.