Also not that there is plenty of code that, in the course of computation, writes into attributes of objects. If you're doing multiple of those it's very hard in python to make such a modification atomic. So if an interrupt ends up splitting an update, you may leave the system in an inconsistent state that could affect future computations. It's a nice aspirational goal to have Ctrl-C work decently, but if you want your results with a reasonably degree of confidence in correctness, you should probably avoid interruptions anyway.
Most code is hopefully written defensively: work on a local copy and only in the last bit modify the attribute by a single assignment. That minimizes the chance for getting inconsistent states from interrupts, but I don't think you'll get it to zero, because of the non-atomicity of multiple attribute assignments. On Wednesday 10 April 2024 at 11:03:13 UTC-7 Vincent Delecroix wrote: > I do not remember anything specific about solve_mod. Though, there are > many places in Cython source code where sig_on/sig_off is not handled > carefully enough (and many that were fixed). > > The fact that it is not reproducible is not necessarily a huge problem. > However, > - Could you share the code (as minimal as possible) that provokes this > on your machine? > - Do you have the complete (C) trace of the crash? You might need to > install gdb depending on your setup. > > Vincent > > On Wed, 10 Apr 2024 at 16:18, Georgi Guninski <ggun...@gmail.com> wrote: > > > > I don't have reproducible testcase, but my pain is that sometimes if I > > press CTL-C e.g. when solve_mod() is called in a loop, I get SEGV and > > abort. > > I suspect besides signals, it is hitting race condition. > > Is this a known issue? > > > > Very long ago there was vulnerability in sendmail, where it did > > printf() or similar in signal handler and this was documented in the > > literature. > > > > -- > > You received this message because you are subscribed to the Google > Groups "sage-devel" group. > > To unsubscribe from this group and stop receiving emails from it, send > an email to sage-devel+...@googlegroups.com. > > To view this discussion on the web visit > https://groups.google.com/d/msgid/sage-devel/CAGUWgD_7t9OTNdo6Ovrt%3DJdNqmGJLPGsZHVoJK0mNPQhuz1CKQ%40mail.gmail.com > . > -- You received this message because you are subscribed to the Google Groups "sage-devel" group. To unsubscribe from this group and stop receiving emails from it, send an email to sage-devel+unsubscr...@googlegroups.com. To view this discussion on the web visit https://groups.google.com/d/msgid/sage-devel/b7d35f4e-7301-4ac4-8cf1-561ba86493e0n%40googlegroups.com.
