[7.1.0] Exclude convenience symlinks after changing the output base

[bazel-io]

When the output base is changed after the convenience symlinks have been created, queries such as //... try to traverse into them and load them as packages since they are only excluded based on whether the resolved location is in the output base. This adds some heuristics to determine if a symlink is a convenience symlink:

1. The symlink name has an appropriate suffix.
2. An ancestor of the symlink target at the appropriate level is called execroot.
3. The execroot directory contains a file called DO_NOT_BUILD_HERE.

These heuristics should work if both the output base and the symlink prefix change while being quite robust to false positives.

This is important for IDE integration, where the output base is often changed for queries, to prevent concurrent builds from blocking them. An example of this is the vscode-bazel extension.

Fixes #10653
Fixes #13951

Closes #21005.

Commit 8f74ace

PiperOrigin-RevId: 610667735
Change-Id: I1869c9a2063f7f526950e48c0b1ee6efa89fd202

[cameron-martin]

It looks like the failing tests on CI are a network issue:

ERROR: Error computing the main repository mapping: in module dependency chain <root> -> bazel_tools@_ -> [email protected]: Error accessing registry https://bcr.bazel.build/: Unknown host: bcr.bazel.build

[iancha1992]

One of the tests is failing now. Could you please take a look? @lberki

cc: @bazelbuild/triage

[cameron-martin]

@iancha1992 this looks like a network issue. See my comment above.

[lberki]

It does look like a network issue. It looks like this has already been merged, but had it not been, I'd have asked you to use sh_library instead of py_library: I could imagine that this test doesn't have network access and thus can't fetch rules_python. But it appears that it does, given that it's been submitted and CI is not very red?

[cameron-martin]

It does look like a network issue. It looks like this has already been merged, but had it not been, I'd have asked you to use sh_library instead of py_library: I could imagine that this test doesn't have network access and thus can't fetch rules_python. But it appears that it does, given that it's been submitted and CI is not very red?

Yeah this passed in the original PR so I'm not sure why this wouldn't. It also passes locally for me so it can't be that networking it disabled via the sandbox.

[cameron-martin]

Seems like this might fix it: #21529

[meteorcloudy]

#21547 should fix the remaining presubmit failures.