[PATCH] D135419: [readobj] Make JSON output consistent for Other flags

[James Henderson via Phabricator via llvm-commits]

jhenderson added a comment.

In D135419#3869971 <https://reviews.llvm.org/D135419#3869971>, @paulkirth wrote:

> Still looking into this, but one thing I'm increasingly hesitant about is how closely the LLVM and JSON implementations are coupled. It may make sense to separate the implementations more before we start introducing changes that affect more than the JSON output.
>
> @jhenderson do we have any stability guarantees for the LLVM format? I don't want to introduce churn just for the sake of it, but is the LLVM output intended to be easily machine parsable, or to follow some format we've outlined somewhere? where changes to both formats make sense, I'm happy to share the implementation, but I expect there will eventually reach a point where the JSON and LLVM formats need to diverge to satisfy consumers. If possible, I'd like to establish where that point is, so that we can plan for/recognize it apriori.

I don't think there's any specific stability guarantees for the LLVM format output, but we'd have to have justifiable reasons to change it. In particular, there are large numbers of tests that use llvm-readobj for testing things, and changing the output may impact those tests. It's possible the changes will be in such a way that a test no longer would fail when a bug was introduced, for example, so you have to be very careful what changes. If you think any non-trivial changes are needed, it would be worth raising on Discourse first too, because other developers may well have a different point of view. In fact, this whole discussion about stability of the output might be worth a Discourse thread, though it would be worth coming up with one or two examples as to what you might want to change in order to make things work.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D135419/new/

https://reviews.llvm.org/D135419