[postbuildlint] Don't read lib info twice #1384
Conversation
@BillyONeal I just realised that you messed this up in #834. Considering that fixing this will most likely need verification with microsoft/vcpkg and e2e tests, how to proceed with this? |
I don't think we've ever seen this break before so I'd be happy to take your fix without new tests. (I would love new tests, but I'm not enough of a macOS person to write them myself and I don't think 'restoring what should have been the status quo' qualifies as the kind of 'new feature' ask for which we require tests.) |
| auto release_lib_info = get_lib_info(fs, release_libs); | ||
| Optional<std::vector<Optional<LibInformation>>> debug_lib_info; | ||
|
|
||
| if (!build_info.policies.is_enabled(BuildPolicy::SKIP_ARCHITECTURE_CHECK) || |
There was a problem hiding this comment.
I'm not a fan of adding conditionals that need to be paired like this or the world ends but I don't have better ideas for now.
|
Do you want me to restore the lib check on OSX in another PR? |
Library information was read twice from the same files: once in
check_lib_architectureand once incheck_crt_linkage_of_libs. This caused a performance regression. This PR moves reading the libraries to a single function (get_lib_info) which is only called once.Additionally,
check_lib_architectureperformed the checkcmake_system_name == "Darwin"but this function is only called on Windows. Therefore, this check is unnecessary and I removed the code inside it.