Print sigfaults and exceptions to STDERR #4163
Conversation
|
Also, while sitting on it, I was wondering does |
src/callstack.cr
Outdated
| if repeated_frame.count == 0 | ||
| LibC.printf "[%ld] %s +%ld\n", repeated_frame.ip, sname, offset | ||
| STDERR.printf "[%d] %s +%d\n", repeated_frame.ip.address, sname, offset |
There was a problem hiding this comment.
These lines use LibC.printf instead of printf in the stdlib to avoid the nonblocking IO code. I'd suggest using LibC.dprintf with fd 2, although you'll have to bind that.
There was a problem hiding this comment.
fprintf expects a FILE*, while dprintf expects only a int fd, so I think the latter would be easier.
There was a problem hiding this comment.
@RX14 thanks for the tip, I've updated the PR accordingly. I was thinking of introducing an enum (STDIN, STDOUT, STDERR) for the fd argument to improve readability, WDYT?
Bind `LibC.dprintf` function
Sija
force-pushed
the
print-sigfault-to-stderr
branch
from
836e651
to
c907c9c
Compare
|
Thanks @Sija! Regarding the pointers, probably the %x is better since they are expected for printing pointers. Feel free to handle that in another PR if you want, but it is not high prio. |
|
I've pushed a followup #4170 PR with that change, cheers! |
While working on raven.cr I've stumbled upon the problem of intercepting exceptions and sigfaults thrown by Crystal code and to my surprise I was unable to do so. After several hours of debugging I've found the problem in using
LibC.printf.This PR changes
__crystal_raise,__crystal_sigfault_handlerandCallStack.print_frameto useLibC.dprintfinstead.