Axel Huebl

thanks for the information, that is super useful! I gave it a little try and indeed in recent versions generateDS depends on lxml with a fallback to ElementTree :) @michaeljones...

Thanks for catching up! :) Unfortunately, the old commands only have a `Generated Mon Feb 9 19:08:05 2009 by generateDS.py.` on top. Newer versions of generateDS add the full command...

> That is part of "step 2", ~~I think the main work now is, after replacing the old generated files with newer ones, to go to manually written files like...

Btw, found the [changelog of generateDS](https://github.com/michaeljones/breathe/files/1721521/README.txt) (probably something like 1.18 in 2009 to 2.29 today.

I have currently added all required `node_type` attributes and am now swiping through side-effects.

I am not fully building, but from a first test the memory consumption is still really bad... https://github.com/ax3l/breathe/tree/topic-elemTree

Thanks for the hint. Yes, the memory consumption is still similarly high, at least ~1.9 GByte for parsing of my ~130 MB doxygen XML input (but it does not yet...

Please everyone, stop spamming this thread and complain directly to your local lenovo support as well as to the people maintaining the linked Ubuntu page. Otherwise, [contribute as described above](https://github.com/nmikhailov/Validity90/issues/34#issuecomment-376230153).

Hi @zombie, yes sorry I forgot to attach one. I was looking at https://github.com/ComputationalRadiationPhysics/picongpu/pull/2016 when I saw it.

Yes, the issue actually more about using Docstring in general for any public API :)