Disable absolute coverage-prefix-map feature by default (#492)

It should only ever be enabled explicitly by tools that require it

Signed-off-by: Aaron Sky <aaronsky@skyaaron.com>

feat: publish stardocs with releases (#474)

Co-authored-by: Aaron Sky <aaronsky@skyaaron.com>

Fix bazel 9.x support (#482)

This is all just around APIs that were removed that we used in tests.
This is copy pasted from bazel tests / rules_apple

Skip post-processing binary with no -o flag (#479)

Since cc_common setup can be called in a way where the output_execpath
is never set, I think we have to support just skipping this.

Fixes #476

Add missing load (#477)

Use args instead of env for output path (#473)

If you have a custom rule that calls cc_common API without passing this
variable, the configuration would fail. In bazel 8.x+ we could add
expand_if_available but we still support 7.x. This uses the same
mechanism as before instead.

Fixes #472

Add missing ZERO_AR_DATE env var (#467)

Somehow I accidentally dropped this in
6e0fdb1 which lead to
#465

The reason we didn't spot this earlier is doing `-Wl,-S` removed the
debug info anyways, so this revealed another issue with that change that
all debug info isn't stripped, but we should have this either way and
can fix that after.

Fixes #465

Revert "Pass -object_path_lto <path> linker flag for LTO builds (#420)…

…" (#426)

Fixes #423

We should re-apply with the actual configs for full_lto and thin_lto, we
should be able to write a test to verify it works too.

This reverts commit 61b44d1.

Added `http_dmg` repository rule for fetching and extracting dmgs (#421)

I've needed this behavior in the past and was able to get something that
works but don't know what the best home would be for such a rule. This
seemed to be the most appropriate place I could find in place of adding.
Looking forward to see if maintainers here agree and if so some
feedback!

Consume the pipes before waiting for pid (#411)

In general one shouldn't call waitpid until the child closed its end of
the pipe. This works fine with libtool invocations that do not produce
more than `_PC_PIPE_BUF` bytes of output but deadlocks when the pipe
fills up.