George Gensure

This is running long. The client is in violation of the spec by identifying a nonexistent input directory, so I'm closing this. If you want to consider an exception (with...

Looking at the protocol, we also have the digest_function field in requests, with the following note: ``` // The digest function the server must use to compute the digest. //...

> * Allow `checksum.sri` to be a list of comma-separated qualifiers with the same meaning (interchangeable hashes, you are allowed to pick one) https://www.w3.org/TR/SRI/#agility refers [to parsing metadata](https://www.w3.org/TR/SRI/#parse-metadata) handling a...

@sluongng > > It would be extremely painful if we do that. > Painful or not, some correspondence with the Digest (and DigestFunction) in the response should be indicated. The...

@sluongng @EdSchouten >> Since the sri only refers to values specified by the client, it sounds like this means that the server must present/compute this digest separately for the purpose...

> In addition, it would be nice to have some other field in the response to signal additional validations performed by the server (like a field specifying which part of...

> That's correct, but right now it's hard to reason about it in the case of `checksum.sri`: If the server validated one out of three checksums that are part of...

> Thanks for the PR, but it still doesn't resolve the issue for me. I'm now seeing NPEs: > Updated this to account for mid-retry null lastException assignment.

We need both as a migration path. An installation maintainer should be able to throw a single config flag (ideally *outside* of buildfarm), and have it just work equivalently. Detection...

Buildfarm builds `//...:all` cleanly with the `--incompatible_auto_exec_groups` flag enabled. Is there some evidence that it does not @sgowroji?