[llvm] [compiler-rt] Reland "[PGO][GlobalValue][LTO]In GlobalValues::getGlobalIdentifier, use semicolon as delimiter for local-linkage varibles. " (PR #75860)

[Mingming Liu via llvm-commits]

minglotus-6 wrote:

> Sometimes, only through rebasing can one notice that the patch needs adjustments to more code or tests due to its interaction with another landed patch

Yeah, in this case there is no merge conflict for the pull request, and the error only manifest after merge happens.  Plus, the presubmit build-bot (for example the [currently pending one](https://buildkite.com/llvm-project/github-pull-requests/builds/23721) for this pull request) takes hours to complete, during which window new changes could in theory come in..

> I am playing with https://getcord.github.io/spr/index.html to use Phabricator workflow without affecting reviewers' experience.

This is helpful to know! Are there known blockers to use `spr` for llvm github review nowadays? [This post](https://discourse.llvm.org/t/stacking-patches-in-github-prs/73403/2) mentions creating branches in upstream (not fork) as one open-ended question but the restriction for branch `users/**/*` was removed [later](https://github.com/llvm/llvm-project/issues/56636#issuecomment-1773583088)

https://github.com/llvm/llvm-project/pull/75860