fix crash at port open

[nowakpl]

Fix for #1771

[ernstnaezer]

I can confirm that this patch fixed my issue with espruino/EspruinoTools#89

[reconbot]

@zbjornson mind taking a look?

[reconbot]

Thank you for this, I haven't had time to dig too deeply.

[zbjornson]

Fix makes sense. Odd that that didn't crash for me. Sorry for the mistake!

It would probably be simpler to keep the struct as it was, and change the call signature of setBaudRate to something like

int setBaudRate(const int fd, const int baudRate, char** errorString)

thereby avoiding the need to create a new instance of the baton at all.

Actually, I'm not sure why a copy of the connection options needs to be created in the first place. Is there a reason we can't just use data? I'm not super familiar with this code yet.

[reconbot]

I think it used the struct to communicate the error, no reason why it couldn't have the string passed in or a different way to do it

[zbjornson]

Just looked more closely -- I was proposing passing a char* by reference for the error data (although I forgot a size parameter earlier). I think that, or changing the hierarchy so that there's a base struct with errorData that all the uv structs inherit from, would be cleaner overall, but will be a bigger commit, so: I'd go with this PR as-is. I might propose a clean-up PR later (but understand if you want to minimize changes to minimize risk too).

I also see now why this didn't crash for me -- there's no equivalent error in serialport_win.cpp.

[reconbot]

I have tested this with some hardware, we're good! thank you so much @nowakpl [email protected] has been released