Richard Levitte

That would be a good thing, methinks > I use clang for webasm builds, it does work there. I can add some guards to try to verify the compiler, though....

A quick detail: ``` + print TAP_FH "%s\n", $tmp->as_string(); ``` [`print`](https://perldoc.perl.org/functions/print) != [`printf`](https://perldoc.perl.org/functions/printf) (I was wondering why your output had all those `%s` all over the place)

I don't understand why PEM files should be treated as binary, anywhere. Can't the buffering issue be handled separately?

On record oriented file systems, like on VMS, using binary mode to read text files (which PEM files are) is often detrimental. On VMS with the default text file record...

I would accept always treating the files on Windows in binary mode, since the impact is "only" that CRLFLF converstion won't take place. So an idea is to replace this...

Okie, I hear and acknowledge your points. Since this is, in the end, a workaround for a workaround of a Windows bug (as shown in the comment), I think that...

This PR has me thinking that an issue is that we call `OSSL_STORE_open_ex()` late in "by store". So, what if we would try to call that early, and store a...

Actually... we might also think of this as a subtle bug in OpenSSL's `file` storemgmt implementation. See, every time that `by_store_subject_ex()`, it's setting a new search criterion, and the `file`...

I was wrong regarding automatic reset. The documentation clearly specifies that OSSL_STORE_find() must only be called before any OSSL_STORE_load(), so a reset will have to be a new feature. Meh...

I thought a bit more about this problem, and realised that #27461 was really only a reset problem in "by store", and could be largely resolved there. So I now...