Expose virtual:original header mappings via CcToolchain variables #21157
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Whenever `prefix` and/or `strip_include_prefix` is passed into `cc_library` rules, Bazel copies the headers into a special "_virtual_includes" directory that allows for such path transformations to occur. Unfortunately, this also has the side-effect of changing how source files are designated within the source binaries: they will refer to the file in the `_virtual_includes` sandbox directory instead of the actual files in the build tree. This change allows us to work around this problem by providing a mapping between the "virtual" and where they are in the filesystem as a [CcToolchainConfigInfo variable][1], `virtual_to_original_dirs`, which ends up being a list of `<virtual>=<original>` mappings that can be used in rules_cc toolchain features, specifically using [-ffile-prefix-map][2]. Questions ========= - Is `<virtual>=<original>` okay, or do we need a more sophisticated lazy expansion involving subvariables like what `libraries_to_link` has? E.g., do we imagine another compiler command-line flag format that'd require `A:B` instead of `A=B`? (Then again, considering MSVC's ASan is implemented w/ `/fsanitize=address`, maybe they're on the `A=B` train now?) Notes ===== - We mostly mimicked coverage's `virtual_to_original_headers`. Fixes upstream bazelbuild#11874. [1]: https://docs.bazel.build/versions/master/cc-toolchain-config-reference.html#cctoolchainconfiginfo-build-variables [2]: https://gcc.gnu.org/onlinedocs/gcc/Overall-Options.html#index-ffile-prefix-map Co-authored-by: Andrew Psaltis <apsaltis@vmware.com>
github-actions
bot
added
team-Rules-CPP
lberki
requested review from
comius
and removed request for
lberki and
oquenchil
comius
reviewed
May 15, 2024
| * context. Unlike {@link virtualToOriginalHeaders}, this is populated even if | ||
| * coverage isn't enabled. | ||
| */ | ||
| private final NestedSet<Tuple> virtualToOriginalDirs; |
There was a problem hiding this comment.
This new field will potentially result in memory regression. There's already virtualToOriginalHeaders, which is probably even larger, containing all header files, however it's only populated with coverage enabled.
I wonder if there's a more efficient way to reconstruct the information from _virtual_includes constant. It would be quite hacky though, but a wrapper script should be able to do it.
comius
added
awaiting-user-response
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Whenever
prefixand/orstrip_include_prefixis passed intocc_libraryrules, Bazel copies the headers into a special "_virtual_includes" directory that allows for such path transformations to occur. Unfortunately, this also has the side-effect of changing how source files are designated within the source binaries: they will refer to the file in the_virtual_includessandbox directory instead of the actual files in the build tree.This change allows us to work around this problem by providing a mapping between the "virtual" and where they are in the filesystem as a CcToolchainConfigInfo variable,
virtual_to_original_dirs, which ends up being a list of<virtual>=<original>mappings that can be used in rules_cc toolchain features, specifically using -ffile-prefix-map.Questions
<virtual>=<original>okay, or do we need a more sophisticated lazy expansion involving subvariables like whatlibraries_to_linkhas? E.g., do we imagine another compiler command-line flag format that'd requireA:Binstead ofA=B? (Then again, considering MSVC's ASan is implemented w//fsanitize=address, maybe they're on theA=Btrain now?)Notes
virtual_to_original_headers.Fixes upstream #11874.