remove precompile mutation step from staticdata #48309
Conversation
vtjnash
added
compiler:precompilation
Make sure things are properly ordered here, so that when serializing, nothing is mutating the system at the same time. Fix #48047
| return NULL; | ||
| // We must avoid attempting to re-enter inference here | ||
| assert(0 && "attempted to enter inference while writing out image"); | ||
| abort(); |
There was a problem hiding this comment.
aborting here in non-debug mode seems a bit aggressive. Generally the image may still be ok even if this fails, and it seems unfriendly to the user to not try and continue.
There was a problem hiding this comment.
There is not really any points in this code that would be calling back into julia now, so I had considered deleting this special flag, but I did want to make it hard for someone to introduce the same mistake again. In most later places, the image might also not be complete, which also seemed more aggravating to encounter for the user than a clear abort and backtrace from the source of the problem
There was a problem hiding this comment.
LGTM. I also favor the assert.
Is the external methodtable the main fix, and most of the rest (nice) cleanup? I like both the conditional- Oh nvm, I now see you're calling edges allocation and the typemap visitor pattern.jl_prepare_serialization_data twice as in #48054 (comment)
Make sure things are properly ordered here, so that when serializing, nothing is mutating the system at the same time.
Fix #48047