[Info-vax] "bad select 38" (OpenSSL on VMS)

[Richard Levitte]

Den fredag 16 september 2016 kl. 15:02:39 UTC+2 skrev Stephen Hoffman:
> On 2016-09-16 12:48:18 +0000, Richard Levitte said:
> 
> > Den fredag 16 september 2016 kl. 13:41:25 UTC+2 skrev Neil Rieck:
> >> My advice you you is to get your working code into the OpenSSL distro 
> >> so that others (including HP/HPE/VSI) aren't required to engage in a 
> >> coding ritual every time a new version of OpenSSL is published.
> > 
> > Done, actually.  The changes finally got merged into the 1.1.0 and 
> > 1.0.2 branches, so whenever 1.1.0a and 1.0.2i are released, mainstream 
> > OpenSSL will finally be free of this particular issue.
> 
> "OpenSSL will finally be free of this particular issue" when OpenVMS 
> has a functional select() call.

Errr, color me confused, but isn't this thread about the issue that OpenVMS currently has a select() that works with sockets only?  Did I miss something?

> When the hack-around you've just added into OpenSSL can be excised.

Sure, as soon as we no longer get stuff like "bad select 38" for trying to use it on fd 0.

> It's these incompatibilities and the associated #ifdef __vms morasses 
> that have helped establish the reputation of OpenVMS in the computing 
> community.
> 
> It's these misfeatures that deservedly and sometimes even gleefully get 
> any latent OpenVMS support and its associated complexity ripped out of 
> open source projects, too.

I agree with you there, partly...  VMS is not Unix, and when someone tries to have Unix as the holy standard, things are sure to get confusing.

That being said, the socket-only select() is lame.

Cheers,
Richard