rfm

> We can shave of another 100-150ms if we reuse change dictionaries. This we need to check carefully (more than I've had chance to). Re-use makes sense but last version...

You were going to address Fred's request for portable testcases. I have done partial fixes recently but afaik there's still quite a bit of broken (ie dependent on Apple/clang specific...

As Hugo says, it appears that clang-format is not currently capably of producing nicely formatted files, with the most significant issue being that it loses the whitespace used to make...

I don't use clang-format, so I don't know exactly what it does right and what it does wrong, but way back when I last looked into automated reformatting one of...

Looking at uncrustify, I'm mot sure how good it is but it sounds like the authors are fairly actively seeking objc improvements, so there may be an opportunity to make...

It's not perfect, but here is an uncrustify config file which seems close. [objc.txt](https://github.com/user-attachments/files/16444648/objc.txt)

Fixed that some time ago, but forgot to close this issue.

What sort of thing did you have in mind? > > On 5 Nov 2024 at 00:39, Hugo Melder ***@***.***> wrote: > > > > It would be great to...

I updated that adding en environment variable setting GS_WITH_ASAN=1 to turn it on. With that it's easy to build both the libraries and run their tests with asan. Of course,...

I spent about 8 days getting most of the base library tests to pass. That's mostly(say 95%) fixing leaks in the testcases (previously there was no incentive to avoid leaking...