#9678: Rewrite interrupt handling
------------------------------+---------------------------------------------
Reporter: jdemeyer | Owner: tba
Type: task | Status: needs_work
Priority: major | Milestone: sage-feature
Component: c_lib | Keywords: interrupt, error, c, cython
Author: Jeroen Demeyer | Upstream: N/A
Reviewer: | Merged:
Work_issues: |
------------------------------+---------------------------------------------
Description changed by jdemeyer:
Old description:
> There are lots of things to be improved in the interrupt handling
> routines in c_lib/src/interrupt.c and c_lib/include/interrupt.h.
>
> Major changes planned:
> * '''DONE''': make {{{sig_on()}}} have function syntax so that we can
> declare it {{{cdef int sig_on() except 0}}}. See #10115 for the syntax
> changes.
> * '''DONE''': do not save signals in {{{sigsetjmp()}}} (by giving a
> second argument of 0 instead of 1). This speeds up a sig_on/sig_off loop
> from 382 clock cycles to 24 clock cycles on a Core(TM)2 Duo CPU T5870 @
> 2.00GHz running Linux 2.6.32 glibc 2.11.2.
> * '''DONE''': using {{{sigaction()}}} instead of {{{signal()}}} since
> that has more well-defined semantics.
> * '''DONE''': handle SIGINT differently from other signals (other
> signals are urgent and cannot be ignored. SIGINT on the other hand does
> not need to handled immediately, but we have to be careful for race
> conditions).
> * '''DONE''': allow {{{sig_on()}}} and {{{sig_off()}}} to be nested.
> * '''IN PROGRESS''': implement {{{sig_retry()}}} for retrying failed
> computations (this is useful for PARI, see #10018).
> * have an interface for more general errors which are not signals.
> * '''DONE''': clean up old, unused code.
> * '''IN PROGRESS''': testing interrupt handling: #10030.
> * '''IN PROGRESS''': fix breakage because of this patch: #10061.
> * '''IN PROGRESS''': documentation: #10109.
> * allow to temporarily suspend interrupts and do this in
> {{{sage_malloc()}}} and friends (an interrupt during {{{malloc()}}} will
> mess up the heap, this happens sometimes with a "fixed" #9564!)
> * '''DONE''': eliminate race condition when a SIGINT arrives before
> {{{sig_on()}}} or during {{{sig_on()}}}.
>
> Other related tickets:
> * #800 (make _sig_on and _sig_off faster when stacked)
> * #9640 (Change PARI error catching mechanism)
> * #9564 (libsingular exponentiation can not be interrupted)
> * #7879 (Remove unnecessary signal handling for low prec mpfr
> operations) --- hopefully {{{sig_on()}}} and {{{sig_off()}}} can be made
> very fast such that this shouldn't be an issue anymore.
> * #10018 (Unhandled SIGSEGV after bnfcertify())
> * #7794 ({{{PolynomialRing_integral_domain}}} ignores Ctrl-C and
> segfaults)
> * #5313 (patch singular so that when it runs out of memory the error
> message says "singular" in it)
> * #3423 (Make Pari error messages more informative)
> * #10126 (Fix error handing in Matrix_rational_dense._invert_pari())
>
> Patch chain: #9893, '''#9678''', #10061, #10030, #10018, #9640
New description:
There are lots of things to be improved in the interrupt handling routines
in c_lib/src/interrupt.c and c_lib/include/interrupt.h.
Major changes planned:
* '''DONE''': make {{{sig_on()}}} have function syntax so that we can
declare it {{{cdef int sig_on() except 0}}}. See #10115 for the syntax
changes.
* '''DONE''': do not save signals in {{{sigsetjmp()}}} (by giving a
second argument of 0 instead of 1). This speeds up a sig_on/sig_off loop
from 382 clock cycles to 24 clock cycles on a Core(TM)2 Duo CPU T5870 @
2.00GHz running Linux 2.6.32 glibc 2.11.2.
* '''DONE''': using {{{sigaction()}}} instead of {{{signal()}}} since
that has more well-defined semantics.
* '''DONE''': handle SIGINT differently from other signals (other signals
are urgent and cannot be ignored. SIGINT on the other hand does not need
to handled immediately, but we have to be careful for race conditions).
* '''DONE''': allow {{{sig_on()}}} and {{{sig_off()}}} to be nested.
* '''IN PROGRESS''': implement {{{sig_retry()}}} for retrying failed
computations (this is useful for PARI, see #10018).
* have an interface for more general errors which are not signals.
* '''DONE''': clean up old, unused code.
* '''IN PROGRESS''': testing interrupt handling: #10030.
* '''IN PROGRESS''': fix breakage because of this patch: #10061.
* '''IN PROGRESS''': documentation: #10109.
* allow to temporarily suspend interrupts and do this in
{{{sage_malloc()}}} and friends (an interrupt during {{{malloc()}}} will
mess up the heap, this happens sometimes with a "fixed" #9564!)
* '''DONE''': eliminate race condition when a SIGINT arrives before
{{{sig_on()}}} or during {{{sig_on()}}}.
Other related tickets:
* #800 (make _sig_on and _sig_off faster when stacked)
* #9640 (Change PARI error catching mechanism)
* #9564 (libsingular exponentiation can not be interrupted)
* #7879 (Remove unnecessary signal handling for low prec mpfr operations)
--- hopefully {{{sig_on()}}} and {{{sig_off()}}} can be made very fast
such that this shouldn't be an issue anymore.
* #10018 (Unhandled SIGSEGV after bnfcertify())
* #7794 ({{{PolynomialRing_integral_domain}}} ignores Ctrl-C and
segfaults)
* #5313 (patch singular so that when it runs out of memory the error
message says "singular" in it)
* #3423 (Make Pari error messages more informative)
* #10126 (Fix error handing in Matrix_rational_dense._invert_pari())
Patch chain: #9893, #9345, '''#9678''', #10061, #10030, #10018, #9640
--
--
Ticket URL: <http://trac.sagemath.org/sage_trac/ticket/9678#comment:21>
Sage <http://www.sagemath.org>
Sage: Creating a Viable Open Source Alternative to Magma, Maple, Mathematica,
and MATLAB
--
You received this message because you are subscribed to the Google Groups
"sage-trac" group.
To post to this group, send email to [email protected].
To unsubscribe from this group, send email to
[email protected].
For more options, visit this group at
http://groups.google.com/group/sage-trac?hl=en.
