[Info-vax] PC/VT Keyboarrd Mapping

[John Reagan]

On Monday, June 27, 2016 at 4:06:21 PM UTC-4, Stephen Hoffman wrote:
> On 2016-06-27 19:33:40 +0000, John Reagan said:
> 
> > Common enough that there was work to make sure the native RTLs and 
> > translated RTLs cooperated.  The Fortran RTLs share LUN information for 
> > example.
> 
> But that's an instance of the "same" RTL translated, and not a 
> completely different RTL?   I'd kind of expect that case to work.

Some of the RTLs that get translated are subtly different with conditional code than the RTLs that ship on the platform.  So in theory, there have been be up to 5 different RTLs that we built (VAX to ship on VAX, VAX to translate to Alpha, Alpha to ship on Alpha, Alpha to translate to Itanium, Itanium to ship on Itanium).  The RTL business isn't for the faint of heart.

> 
> Now if I had code using VSI Clang LLVM RTL C11, I'm not so sure I'd 
> expect that to interoperate with other translated code using a 
> translated VAX C RTL and native Compaq C RTL circa C90 mixed in for the 
> most complete code conflagration.

Those are the kind of scenarios that need to be discussed.  I'm all for breaking stuff. :) 
> 
> Skewing somewhat away from blanket upward-compatibility and making new 
> code easier to write and to maintain is preferable, though that'll be 
> unpopular with some existing code in the short term, but offset by 
> better support, features and stability with the newer RTL over the mid- 
> and longer-term.   As an example of this stalled migration, getting 
> folks off of VAX C should not still be going on.  Folks can either fix 
> the ancient application code, or stay on the ancient release where that 
> code best belongs.   Do I like fixing crufty old code?  No.  But 
> getting that old code off VAX C made the code vastly more stable.
> 
> > I'm guessing that most folks really don't want to toss ALL the baggage. 
> >  Go and do
> > 
> > $ DEFINE/SYSTEM/EXEC DECC$UNIX_LEVEL 90
> > 
> > and sit back and wait for the phone calls to start. I don't think even 
> > the C compiler will work with that turned on.
> 
> Tried that knob some years ago.   Those switches are much like playing 
> minesweeper.  Various routines blow up within (formerly) running code.  
> That's also part of why I utterly despise that mechanism, as some of 
> those errors can be quite subtle when you're off app-stacking and 
> somebody wants one of those knobs and somebody else has an adverse 
> reaction.  But then I (again) realize that basename() is utterly borked 
> when passed OpenVMS filenames, and wonder why I even bother.
> 

I'll have somebody take a run at basename().  It does work for the simple cases that most people have.  Send me your edge conditions and I'll add them to the list.  I actually think it is time for me to start another thread to talk about the CRTL.