Fix sig_on/sig_off bugs

[jdemeyer]

Thanks to #10030, we can check for invalid uses of sig_on()/sig_off(). This ticket will collect those bugs (the bugs do not depend on #9678 or #10030, they were just found using those tickets).

Patch chain: #9678, #10061, #10258, #10030, #10018, #9640

CC: @vbraun

Component: c_lib

Keywords: interrupt

Author: Jeroen Demeyer

Reviewer: Volker Braun

Merged: sage-4.7.alpha1

Issue created by migration from https://trac.sagemath.org/ticket/10061

[jdemeyer]

Changed keywords from none to interrupt

[jdemeyer]

Author: Jeroen Demeyer

[vbraun]

comment:8

Looks good to me.

We (i.e. you) should document somewhere very clearly the use of

   sig_on()
   c_function()
   sig_off()

vs.

   try:
      sig_on()
      cython_function()
   finally:
      sig_off()

In any case, positive review to this ticket.

[jdemeyer]

Reviewer: Volker Braun

[jdemeyer]

comment:9

There is already some documentation at #10109, but you are right that it should be expanded.

[jdemeyer]

Attachment: 10061_fix_sig_on_sig_off.patch.gz

[jdemeyer]

comment:10

Rebased patch

[vbraun]

comment:11

What did you rebase it against? Fails to apply on sage-4.6.2.alpha3

[jdemeyer]

comment:12

Replying to @vbraun:

What did you rebase it against? Fails to apply on sage-4.6.2.alpha3

The future sage-4.6.2.alpha4 :-)

For this patch, that probably means sage-4.6.2.alpha3 + #10028.

[jdemeyer]

comment:14

Additional patch needs review.

[jdemeyer]

Additional patch

[vbraun]

comment:16

Attachment: 10061_newforms.patch.gz

The additional patch looks good to me, too.

[jdemeyer]

Merged: sage-4.7.alpha1