I think having a reason is a good idea, and I can see the advantage if they were descriptive, but if it's just an identifier can't you get that information anyway from -G:+TruffleExceptionsAreFatal and -G:+DumpOnError?

What do @nirvdrum and @eregon think?

@chrisseaton I think with these options I could only trace the first optfail that occurs.

👍 for adding the reason, but I think a descriptive comment is quite more helpful and less cryptic.
I do see the point of having immediately a unique identifier to find the source though, Graal/Truffle should really provide that if possible.
A good description of the reason would be quite nice as a documentation to why that method is not designed for compilation since it is not always trivial.

@eregon I agree something descriptive would be more useful (still should be a unique message though).
I only used a UUID right now because it would take significant amount of time to write a description for the 500+ places notDesignedForCompilation is used.

Maybe we could just derive some description from the node class name or the particular specialization node? Or use the backtrace if the Truffle description with -G:+DumpOnError is not precise enough.

@eregon also 👍 , I wish this was provided: "Graal/Truffle should really provide that if possible."

Maybe I'm missing something, but if Truffle auto-generated the ID, how would you grep through the source for it?

@eregon I tried using the node to get a backtrace but for some reason this was causing Graal to just hang when run

@nirvdrum I think he's thinking something like a backtrace line number

This is the script I used to update these

require 'securerandom'
files =  Dir.glob("truffle/**/*.java")
puts files.size
puts files.to_s
files.each do |filepath|
   text = File.read(filepath)
   replace = text.gsub(/notDesignedForCompilation\(\);/)  { |num| "notDesignedForCompilation(\"#{SecureRandom.uuid.gsub('-','')}\");" }
   File.open(filepath, "w") {|file| file.puts replace}
end

@bjfish do you have a particular example at hand so we can compare the alternatives?

@bjfish That's us generating the ID, which I'm actually okay with.

@eregon Sorry, I don't think I have anything to compare.

The dynamic stuff I've tried just causes Graal to hang / be really slow.

@bjfish we really appreciate this PR but we don't think it's right for the project. I think the unique identifiers in the source code look really out of place and really the emphasis should be on removing these notDesignedForCompilations rather than finding ways to making working with them easier.

We do understand your use case though, and we'll look at adding an option to Graal that prints the backtrace that you get with TruffleCompilationExceptionsAreFatal but doesn't actually exit.

Since you have your script, you can actually keep running that locally and then just not commit the results if you want to.