[PWGJE] Add PCH for JET tasks

[singiamtel]

Heavily speeds up compilation

CC @ktf

[github-actions]

O2 linter results: ❌ 1 errors, ⚠️ 0 warnings, 🔕 0 disabled

[fjonasALICE]

what is the time difference in compilation that you observe? should we consider this for other JE tasks too? also, are there some other headers that do not frequently change in O2Physics that we should consider precompiling?

[ktf]

@singiamtel very nice. Could you share numbers on the improvement? How does it handle changes in the framework headers? Will it actually notice and recompile? That's where the previous attempt I had did not converge.

[aalkin]

This is great, together with my changes at the framework level this will reduce the compilation times very significantly. However, I suggest restructuring this to make it more transparent for the end users, we do not want to rewrite every CMakeLists.txt in O2Physics. We can introduce a convention on the "central" target for each PWG, that collects all the universal headers and then update o2physics_add_dpl_workflow to automatically use it.

[ktf]

@vkucera on a separate note, could we please make sure the O2 linter does not tell people to change QA to Qa ?

[ktf]

@aalkin, as discussed offline and as I think we agree now, I think this approach (coupled with #14668) is actually better. We do need to have the common precompiled headers defined explicitly and to have the users select the appropriate one, like they do for dependencies.

[vkucera]

@vkucera on a separate note, could we please make sure the O2 linter does not tell people to change QA to Qa ?

Why?

[ktf]

Because the abbreviation of Quality Assurance is QA, not Qa.

[vkucera]

Because the abbreviation of Quality Assurance is QA, not Qa.

Yet we have McCollisions and McParticles, not MCCollisions and MCParticles, as names of tables and we have alice3-pid-rich-qa and hf-task-mc-validation, not a-l-i-c-e-3-p-i-d-r-i-c-h-q-a and h-f-task-m-c-validation, as names of workflows.

Capitalising only the first letter of each word-like segment is the only way to keep the CamelCase convention for structs, files, and types consistent with the kebab-case convention for workflows and devices.

[ktf]

I am not sure I understand. Can't you simply add the exceptions? There is hopefully a finite number of them. If the tools are wrong, we should fix the tools, or at least remove the check.