Thomas Stuefe

> The gist of this patch is code like this: > > ``` > jlong CgroupV1Subsystem::read_memory_limit_in_bytes() { > GET_CONTAINER_INFO(julong, _memory->controller(), "/memory.limit_in_bytes", > "Memory Limit is: " JULONG_FORMAT, JULONG_FORMAT, memlimit); >...

I think the gist of my remark is that I would like the layers to behave consistently. I see that `CgroupSubsystem::memory_limit_in_bytes()` is only used in two places, `os::Linux::available_memory() ` and...

> > > > The gist of this patch is code like this: > > > ``` > > > jlong CgroupV1Subsystem::read_memory_limit_in_bytes() { > > > GET_CONTAINER_INFO(julong, _memory->controller(), "/memory.limit_in_bytes", >...

> > I think the gist of my remark is that I would like the layers to behave consistently. > > I see that `CgroupSubsystem::memory_limit_in_bytes()` is only used in two...

> Hi @tstuefe , > > Thanks for the detailed explanation (and diagrams!) of your concern about the complexity of this. I understand what you mean. This PR was my...

> Basic moving of the parsing functions to their own header file (modulo the naming of that file) seems okay. But I don't follow the "memory size" aspect of this...

Hi @dholmes-ora, @stefank, thanks for looking at this. I worked in your feedback. @dholmes-ora: I hope my comment above made my intent a bit clearer. Cheers, Thomas

BTW, this fix intends to simplify the solution for https://github.com/openjdk/jdk/pull/10144 which then will make it possible to remove -XX:MallocMaxTestWords and its supporting code, and use NMT malloc limits instead (e.g....

> > Before we had: > > A) parse_integer() -> does not parse integer, but memory sizes, e.g. "1K", "1G", etc > > Hmmm, I don't consider those "memory sizes"...

> I'm not sure why you need to have two different parsing routines when one routine can parse everything. Things like 1K, 1M etc are defined by Unicode as "compact...