Tiago Quelhas

Yes, we can include it in the BES. I'll prepare a PR.

df07d275ce17f14224af0f1a5862755a7db44397 touched some code around this area, so you'll have to rebase your PR.

@bazel-io fork 7.1.0

If it's an unresolved symlink, I'm not even sure we should do any path transformation at all. That has the potential to break cases where the symlink target path is...

I like this idea too! But to be clear, the new behavior of `--incompatible_sandbox_hermetic_tmp` wouldn't be exactly the same as `--sandbox_add_mount_pair=/tmp`, right? AIUI, the latter not only makes the system...

I really don't like the complexity entailed by resolving symlinks; it's always more complicated than it seems. What if, for example, an output is materialized as a chain of symlinks...

At this time, we'd prefer to avoid making the download logic header-dependent (because it opens the door for doing the same for other HTTP headers, in the limit bringing the...

Thanks for the additional context! However, I maintain my opinion that the best way forward is to add Range support to the Remote Asset API.

A repro would help us make progress on this. Is anyone able to provide it?

I ran `bazel info peak-heap-size used-heap-size max-heap-size used-heap-size-after-gc` in a loop with `--remote_executor` set, but I couldn't reproduce it. There's probably some other flag that need to be set? Please...