The-OpenROAD-Project
Add `tclint` to OpenROAD CI
Description
This issue is a follow-up from discussion on #4337, and tracks adding tclint to the OpenROAD CI.
tclint is a lint tool for Tcl that checks both style and functional errors. I've done some early experiments on this codebase, and with a baseline config there are ~23,000 lint violations across all the Tcl scripts. Nearly all of these are style-related.
I think the best strategy would be to add linting incrementally to certain parts of the codebase. This will also give the maintainers a chance to review a reasonably-sized initial PR to make sure the style enforced by tclint is to their liking.
Do any maintainers have a preference for a chunk of the Tcl codebase that might be a good candidate for the first linting pass? If given a pointer somewhere I'll open a PR that adds linting to CI for that part of the code and does the cleanup required to get it to pass.
Suggested Solution
No response
Additional Context
No response
Should create a mirrored issue for ORFS.
On that note, it might be good to create separate issues for a Python linter as well.
@nmoroze are you planning on enhancing it reformat files similar to clang-format -i? That would be super useful.
Should create a mirrored issue for ORFS.
Done: https://github.com/The-OpenROAD-Project/OpenROAD-flow-scripts/issues/1686
are you planning on enhancing it reformat files similar to clang-format -i? That would be super useful.
yeah, this was on my medium/long-term roadmap but I can consider prioritizing it, since it definitely seems useful for getting lint going on a large existing codebase like this one. At the very least, I can do some early experiments to see how hard it will be. I think tclint has most of the key pieces already, and "just" needs an emitter for the syntax tree to spit out well-formatted Tcl scripts. I tried to be mindful of this use case when designing the parse tree data structure to make sure it captures enough information, but the devil's in the details and I have a feeling there will be some tricky bits :)
I opened an issue to track this: https://github.com/nmoroze/tclint/issues/8
I've submitted the first PR (#4397) towards this effort, cleaning up Resizer.tcl as suggested.
One facilitating change to tclint was adding an option to not require an extra level of indentation within namespace eval blocks - this seems to be a common pattern in the OpenROAD code, and it seems to halve the overall number of lint violations found in the codebase.
If that PR looks good, I can go ahead and clean up the entire src/rsz directory unless anyone has alternate suggestions (which will hit a bunch of Tcl files in src/rsz/test/). There are ~200 violations there, so it's chunkier but tractable.