Closed davidedelvento closed 1 week ago
@wsmoses maybe this problem could be due to infinite allocation in the stack?
@wsmoses did you have any time to look at this issue?
I have not unfortunately, do you want to find a time later this week or next to look at togther (it should be a fast fix, but might as well confirm it works in your integration tests together)
@davidedelvento pardon the delay, but does this fix it? https://github.com/EnzymeAD/Enzyme/pull/1977
Closing per persumed fix, reopen if persists
This https://github.com/EnzymeAD/Enzyme/pull/1983 fixes is. Thanks a lot!
@wsmoses , this is on the
origin/gsl
branch which I am attempting to verify with the following test case.Note: this is likely to be "operator error" in that I am using Enzyme incorrectly, but I am still reporting it as a "bug" in that it should refuse to compile rather than cause a crash at runtime.
following the footsteps of https://github.com/EnzymeAD/Enzyme/issues/792#issuecomment-1224701354
To compile, I use the following Makefile
Running the test code prints everything, but segfaults (and hence finishes with exit code 139, despite the "return 0")
Removing the
__enzyme_autodiff
call makes the code work as expected (obviously without the wanted results :rofl: )Running the print option with
make debug
provides the following, which I am not fluent enough yet to understand deeply enough to grasp what the problem might be.Replacing
print
withprint-activity
yields: