[llvm] [symbolizer] Empty string is not an error (PR #92660)
via llvm-commits
llvm-commits at lists.llvm.org
Thu May 23 11:14:02 PDT 2024
pcc wrote:
> Let's go with this then if there aren't any objections? To be clear, I'm happy with either a line starting with ECHO being completely echoed back to the client, or the ECHO being stripped off and the rest being echoed back. Whichever approach we go with, the feature should be documented in the llvm-symbolizer docs, which should include a use-case or two explaining why you might use this (e.g. to establish that llvm-symbolizer is ready, or to insert comments into the output stream).
Thanks, either approach works for me, with a slight preference for echoing the whole line back including the `ECHO`.
> Compatibility is an important motivation, if we can maintain it for wider set of cases, we probably should. However if compatibility is your primary concern, the responce to empty string must be empty string, as this is the behavior of older versions. More complex patterns, like sending some non-empty commands don't address your concern, do they?
My primary concern is that there should be a way to do this. I don't have a particularly strong opinion on *how* we do it, but it seems preferable if we can do it in a way that's backwards compatible. By that I mean that it should be possible for me to write my program so that it accepts the behavior of new versions of llvm-symbolizer as well as older versions. For example, if we adopt either variant of the `ECHO` proposal and my program writes `ECHO\n` and expects one of `ECHO\n` or `\n` then it will be compatible with new versions of llvm-symbolizer as well as pre- https://github.com/llvm/llvm-project/commit/1792852f86dc75efa1f44d46b1a0daf386d64afa versions.
https://github.com/llvm/llvm-project/pull/92660
More information about the llvm-commits
mailing list