Migrate from toolchains_llvm to llvm

[jwnimmer-tri]

Extracted from #24810.

Closes #24151.

This no longer lets us independently choose the clang-format version from the bazel_dep module version, but that should be fine.

---

This change is 

[jwnimmer-tri]

+a:@tyler-yankee for feature review or delegation, please.
+a:@rpoyner-tri for platform review & build SME review, please.

[tyler-yankee]

Just some clarifying questions.

@tyler-yankee reviewed 5 files and all commit messages, and made 4 comments.
Reviewable status: 3 unresolved discussions, LGTM missing from assignees rpoyner-tri(platform),tyler-yankee, commits need curation (https://drake.mit.edu/reviewable.html#curated-commits) (waiting on jwnimmer-tri and rpoyner-tri).

---

MODULE.bazel line 271 at r1 (raw file):

    name = "llvm",
    dev_dependency = True,
    llvm_versions = {"": "22.1.7"},

BTW if we care, it seems https://registry.bazel.build/modules/llvm/ does allow what the PR description claims to lose - under "Selecting the LLVM version":

llvm = use_extension("@llvm//extensions:llvm.bzl", "llvm")
llvm.version(llvm_version = "22.1.0")

but would this only be viable if you're actually using the whole toolchain (and not a local one)?

---

MODULE.bazel line 273 at r1 (raw file):

    llvm_versions = {"": "22.1.7"},
)

It's not clear to me why we don't have/need(?) a register_toolchains, use_repo, etc. with llvm.

ah, because we're actually using a local toolchain, and this is only pulling in the tools that we grab elsewhere? I could buy that.

---

MODULE.bazel line 262 at r1 (raw file):

bazel_dep(
    name = "llvm",
    version = "0.8.9",

BTW why not 0.8.16? Bazel version incompatibility, newer clang-format (though, see other comment), something else?

[jwnimmer-tri]

@jwnimmer-tri reviewed 5 files and all commit messages, made 3 comments, and resolved 2 discussions.
Reviewable status: 1 unresolved discussion, LGTM missing from assignees rpoyner-tri(platform),tyler-yankee, commits need curation (https://drake.mit.edu/reviewable.html#curated-commits) (waiting on rpoyner-tri and tyler-yankee).

---

MODULE.bazel line 262 at r1 (raw file):

Previously, tyler-yankee (Tyler Yankee) wrote…

BTW why not 0.8.16? Bazel version incompatibility, newer clang-format (though, see other comment), something else?

Ah, I didn't actually look closely, this was just the original PR text. Bumped to latest now.

---

MODULE.bazel line 271 at r1 (raw file):

Previously, tyler-yankee (Tyler Yankee) wrote…

BTW if we care, it seems https://registry.bazel.build/modules/llvm/ does allow what the PR description claims to lose - under "Selecting the LLVM version":

llvm = use_extension("@llvm//extensions:llvm.bzl", "llvm")
llvm.version(llvm_version = "22.1.0")

but would this only be viable if you're actually using the whole toolchain (and not a local one)?

Right. Configuring the model extension doesn't change the version of the particular @llvm//tools:... aliases we're using.

---

MODULE.bazel line 273 at r1 (raw file):

Previously, tyler-yankee (Tyler Yankee) wrote…

It's not clear to me why we don't have/need(?) a register_toolchains, use_repo, etc. with llvm.

ah, because we're actually using a local toolchain, and this is only pulling in the tools that we grab elsewhere? I could buy that.

Exactly.

Any suggestions on documentation that would have helped?

[tyler-yankee]

@tyler-yankee reviewed 1 file and all commit messages, and made 2 comments.
Reviewable status: 2 unresolved discussions, LGTM missing from assignees rpoyner-tri(platform),tyler-yankee, commits need curation (https://drake.mit.edu/reviewable.html#curated-commits) (waiting on jwnimmer-tri and rpoyner-tri).

---

a discussion (no related file):

Closes #24151.

Does llvm have better backing for its downloads, or? Is it still downloading and extracting the whole LLVM project?

---

MODULE.bazel line 273 at r1 (raw file):

Previously, jwnimmer-tri (Jeremy Nimmer) wrote…

Exactly.

Any suggestions on documentation that would have helped?

Looking up we already have

# Load developer dependencies that are modules.

for this section, which I had originally missed, and certainly doesn't hurt.

WDYT?

# Obtain LLVM for its development tools only. We don't register its toolchain
# because our local C/C++ toolchain is already configured above.

[jwnimmer-tri]

@jwnimmer-tri reviewed 1 file and all commit messages, and made 2 comments.
Reviewable status: 2 unresolved discussions, LGTM missing from assignees rpoyner-tri(platform),tyler-yankee, commits need curation (https://drake.mit.edu/reviewable.html#curated-commits) (waiting on rpoyner-tri and tyler-yankee).

---

a discussion (no related file):

Previously, tyler-yankee (Tyler Yankee) wrote…

Closes #24151.

Does llvm have better backing for its downloads, or? Is it still downloading and extracting the whole LLVM project?

The downloads are smaller, but are still served single-source: https://github.com/hermeticbuild/hermetic-llvm/releases/tag/llvm-22.1.7-2. I figured it closes the issue because at least if nothing else the failure mode (/ error message) is going to be different, as well as any approach we'd use to fix the code.

---

MODULE.bazel line 273 at r1 (raw file):

Previously, tyler-yankee (Tyler Yankee) wrote…

Looking up we already have

# Load developer dependencies that are modules.

for this section, which I had originally missed, and certainly doesn't hurt.

WDYT?

# Obtain LLVM for its development tools only. We don't register its toolchain
# because our local C/C++ toolchain is already configured above.

Done.

[tyler-yankee]

feature thanks!

@tyler-yankee reviewed 1 file and all commit messages, made 1 comment, and resolved 2 discussions.
Reviewable status: LGTM missing from assignee rpoyner-tri(platform), commits need curation (https://drake.mit.edu/reviewable.html#curated-commits) (waiting on rpoyner-tri).

[rpoyner-tri]

@rpoyner-tri reviewed 5 files and all commit messages, and made 1 comment.
Reviewable status: commits need curation (https://drake.mit.edu/reviewable.html#curated-commits) (waiting on jwnimmer-tri).

[rpoyner-tri]

+(status: squashing now)

@rpoyner-tri made 1 comment.
Reviewable status: complete! all discussions resolved, LGTM from assignees rpoyner-tri(platform),tyler-yankee (waiting on jwnimmer-tri).