Details
- Reviewers
benlangmuir - Commits
- rGbeebad9a9b23: [clang][deps] NFC: Remove dead code
Diff Detail
- Repository
- rG LLVM Github Monorepo
Event Timeline
Removing ImplicitModulePCMPath LGTM; not sure about the other change.
clang/lib/Tooling/DependencyScanning/ModuleDepCollector.cpp | ||
---|---|---|
178 | I see we're enabling strict hashing in the scaner itself: ScanInstance.getHeaderSearchOpts().ModulesStrictContextHash = true;, which makes me think this code was never used to influence the scanner's implicit build. If that's true, was this code *already* dead before my change in D129884? It's not clear to me what it was doing. |
clang/lib/Tooling/DependencyScanning/ModuleDepCollector.cpp | ||
---|---|---|
178 | The line you found enables strict context hashing for the CompilerInstance that performs the scanning build. That's still important so that we don't squash multiple module configurations into one minimized PCM. The line this patch removes used to control generation of the module context hash we report to the client. This used to be done by taking the original TU command line, tweaking it, and calling CompilerInvocation::getModuleHash(). Since we now hash the whole command line that's generated from said CompilerInvocation, controlling the behavior of getModuleHash() by enabling strict context hash is no longer necessary. |
clang/lib/Tooling/DependencyScanning/ModuleDepCollector.cpp | ||
---|---|---|
178 | Got it, thanks for explaining! |
I see we're enabling strict hashing in the scaner itself: ScanInstance.getHeaderSearchOpts().ModulesStrictContextHash = true;, which makes me think this code was never used to influence the scanner's implicit build. If that's true, was this code *already* dead before my change in D129884? It's not clear to me what it was doing.