[workspace] Switch conex to use hidden namespace

[jwnimmer-tri]

Towards #17231.

See also ToyotaResearchInstitute/conex#6.

---

This change is 

[jwnimmer-tri]

+@xuchenhan-tri for feature review, please.

[jwnimmer-tri]

Reviewable status: 1 unresolved discussion, LGTM missing from assignee xuchenhan-tri(platform), needs at least two assigned reviewers

---

multibody/contact_solvers/conex_supernodal_solver.cc line 7 at r1 (raw file):

#include "drake/common/drake_export.h"
#include "drake_vendor/conex/clique_ordering.h"

Working

I might try not adding the drake_vendor prefix to see if it works better.

[jwnimmer-tri]

Reviewable status: LGTM missing from assignee xuchenhan-tri(platform), needs at least two assigned reviewers

---

multibody/contact_solvers/conex_supernodal_solver.cc line 7 at r1 (raw file):

Previously, jwnimmer-tri (Jeremy Nimmer) wrote…

Working

I might try not adding the drake_vendor prefix to see if it works better.

Done

[xuchenhan-tri]

Feature
+@rpoyner-tri for platform review please.

Reviewed 5 of 8 files at r1, 3 of 3 files at r2, all commit messages.
Reviewable status: 4 unresolved discussions, LGTM missing from assignee rpoyner-tri(platform)

---

tools/workspace/BUILD.bazel line 49 at r2 (raw file):

    srcs = ["vendor_cxx.py"],
    visibility = [
        # These should all be of the form "@foo_internal//:__subpackages__".

BTW,

---

tools/workspace/conex_internal/vendor.bzl line 16 at r2 (raw file):

    For "supernodal_solver", declares a cc_library in the usual sense, except
    rewrites the source code so that it uses a `drake_vendor` inline hidden
    namespace wrapper so that the library's linker symbols are private.

BTW, understanding check, the namespace wrapper is applied by vendor_cxx and this bazel function simply routes to cc_library_vendored?

Code quote:

    rewrites the source code so that it uses a `drake_vendor` inline hidden
    namespace wrapper so that the library's linker symbols are private.

---

tools/workspace/conex_internal/vendor.bzl line 49 at r2 (raw file):

    ]
    edit_include = None
    strip_include_prefix = "/conex/drake_vendor"

BTW, I don't understand how this works. Don't the vendored sources and headers have prefix drake_vendor/conex?

Code quote:

strip_include_prefix = "/conex/drake_vendor"

---

tools/workspace/conex_internal/patches/eigen_namespace_pollution.patch line 1 at r2 (raw file):

[conex] Fix RLDLT to live in the conex namespace

BTW, is there a path towards merging the PR upstream? Do we really need the patch?

[jwnimmer-tri]

Reviewable status: 2 unresolved discussions, LGTM missing from assignee rpoyner-tri(platform)

---

tools/workspace/BUILD.bazel line 49 at r2 (raw file):

Previously, xuchenhan-tri wrote…

BTW,

Done?

---

tools/workspace/conex_internal/vendor.bzl line 16 at r2 (raw file):

Previously, xuchenhan-tri wrote…

BTW, understanding check, the namespace wrapper is applied by vendor_cxx and this bazel function simply routes to cc_library_vendored?

Yes. I've amended the comment a bit, too.

---

tools/workspace/conex_internal/vendor.bzl line 49 at r2 (raw file):

Previously, xuchenhan-tri wrote…

BTW, I don't understand how this works. Don't the vendored sources and headers have prefix drake_vendor/conex?

The semantics of this string differ depending on whether or not it opens with a /. It will probably be clearest, though, to use the spelling without the slash.

---

tools/workspace/conex_internal/patches/eigen_namespace_pollution.patch line 1 at r2 (raw file):

Do we really need the patch?

Yes. Otherwise we end up vendoring the class as drake_vendor::Eigen::RLDLT which doesn't work at all. (It can't find the real Eigen symbols.)

Is there a path towards merging the PR upstream?

See the link below on line 7. I've submitted this patch upstream as a pull request.

From past experience, though, upstream is dead. There are trivial open PRs there that are 6+ months old with no reply.

[xuchenhan-tri]

Reviewed 1 of 1 files at r3, all commit messages.
Reviewable status: 1 unresolved discussion, LGTM missing from assignee rpoyner-tri(platform)

---

tools/workspace/BUILD.bazel line 49 at r2 (raw file):

Previously, jwnimmer-tri (Jeremy Nimmer) wrote…

Done?

I was able to figure this one out but forgot to deleted the half finished thought. Apologies.

---

tools/workspace/conex_internal/vendor.bzl line 49 at r2 (raw file):

Previously, jwnimmer-tri (Jeremy Nimmer) wrote…

The semantics of this string differ depending on whether or not it opens with a /. It will probably be clearest, though, to use the spelling without the slash.

Oh I see. Thanks for the reference.

[rpoyner-tri]

Reviewed 5 of 8 files at r1, 2 of 3 files at r2, 1 of 1 files at r3, all commit messages.
Reviewable status: complete! all discussions resolved, LGTM from assignees rpoyner-tri(platform),xuchenhan-tri(platform)

[ggould-tri]

FYI, this PR is the primary suspect in two CI failures,

• linux-focal-gcc-bazel-nightly-coverage/1164
• linux-focal-clang-bazel-nightly-coverage/1159

At @svenevs request we (#buildcop) are holding off on reverting this in hope of a fix-forward.