I've had a play around with c-ares today. I've added some code to get the DNS servers (which is what I thought the problem was) but it hasn't helped.

I cheated a bit by only opening bsdsocket for this one function - probably it needs to use it throughout the code rather than clib/newlib (especially when it gets linked to curl as that uses bsdsocket directly), but I don't think that's the problem here.

If anybody wants to look at it my fork is here:
https://github.com/chris-y/c-ares/tree/amiga

@MickJT

The AmiSSL one doesn't have the threaded resolver enabled - for whatever reason it doesn't work.

I'm going to have a look at c-ares as I think I know what's wrong with it.

@redfox

The slowness will be the AmiSSL build doesn't use the threaded resolver. I wouldn't expect it to be *that* slow though.

Need to work out how to fix the threaded resolver, or get libcares working.

@Maijestro

This is odd, I'd expect this behaviour with gethostbyname_r as per salass00's earlier comments.

Does it happen if you stick to non-https sites?

@Maijestro

When it eventually timed out did you get an error message? Can't see in the video as it processes the window close so any error doesn't stay on screen.

Hmm....

OK, this one uses AmiSSL but not gethostbyname_r():

https://www.cy2.uk/netsurf_os4_amissl_no_r.lha

Hmm, OK, I need to go back to AmiSSL but with this new patch then!

Curl uses gethostbyname_r in the threaded and other resolver, so with this patch it should stop it using it regardless.

I'd like to get the threaded resolver working again though.

@boabster
You can use something like pastebin, but at this stage unless you're seeing some other problem it's not going to tell me anything I don't already know.

NEW BUILD: https://www.cy2.uk/netsurf_os4_no_r.lha

This one reverts back to OpenSSL but removes any use of gethostbyname_r.

Hopefully this one will work for everybody!

I tried c-ares again and it doesn't resolve anything.

I've tracked my original error down to init_resolve_thread() failing. I don't see why that would fail if I build against AmiSSL but be OK if I don't! Very odd.

@joerg

The NetSurf build of libcurl uses c-ares on some platforms. I did try it on OS4 but it didn't work (can't remember what the problem was, maybe just didn't resolve anything?)

That's interesting but not sure it's the problem here as I had to disable the threaded resolver to get it to work at all. It's possible that it didn't actually disable though.

@Maijestro

If you look in users/(your username)/ in the NetSurf dir, there's a file called "Choices". Add the suppress line to that.

(your username) - you probably only have one directory here, it might be called Default.

@Maijestro

Ok, that's just showing Curl timeout. Next thing to try is to add this to Choices:
suppress_curl_debug:0

That should show what Curl is doing.

@Maijestro

Can you use that newer zip archive? and I only want the output when it crashes - as it should reveal which setting is causing it.

@Maijestro

Can you use this version ==> https://www.cy2.uk/ns_os4_dbg.zip (just overwrite the executable)

Run it with netsurf -v *>ram:ns.log and send the log to me, or post the last few lines here.

Thanks!

@Maijestro

Thanks, I think there's probably a double free or a mismatched alloc/free. Can you run with netsurf -v and see if it dumps the options as it frees them? I can't remember if it does, but I need to figure out which option is causing the crash.

@Maijestro

Simple crashlogs are fine - all I did was enable line debugging so I can pinpoint the problem.

Debug build: https://www.cy2.uk/netsurf_os4_amissl_debug.lha

Anybody with crashlogs needs to use this version please.

@Boabster

Can you get a crash log over serial? (I know it's a pain to do)

I might need to rebuild with the debug symbols back in to pinpoint it though.

@Boabster

Hmm, can you make sure you're on the latest AmiSSL.