[cling] Use single Parser for LookupHelper

[hahnjo]

It is the only construction of a temporary parser, and it seems not necessary (anymore).

---

This will allow us to drop some downstream Clang patches; not included yet, will need some sequencing with other changes in our monorepo fork. Let's see what the CI has to say first 🙃

[github-actions]

Test Results

    17 files      17 suites   4d 3h 3m 28s ⏱️
 2 680 tests  2 680 ✅ 0 💤 0 ❌
43 906 runs  43 906 ✅ 0 💤 0 ❌

Results for commit 399c950.

♻️ This comment has been updated with latest results.

[vgvassilev]

Probably worth checking with cmssw.

[hahnjo]

Probably worth checking with cmssw.

Yes, good idea, first wanted to see if our own CI agrees with my local testing. I'll allow CMS people to come back from vacation first and also update to ROOT master before asking.

[pcanal]

The original reason for the separate parser (as seen in its introduction: b1003a1) is to side-steps potential failures/bugs in the parser error recovery.

[hahnjo]

Ok, but I'm not sure if a second parser is actually sufficient for that: In case of problems with the parser error recovery, the main parser for Cling is untouched and continues to work, but now the single parser used by the LookupHelper is corrupt and future queries to it will fail for unrelated reasons. If we created a new temporary parser for every lookup, the story would be different, but with the current state what am I missing?