RDI on OpenBSD

RDI on OpenBSD

[Daniel Néri]

[-- Attachment #1: Type: text/plain, Size: 261 bytes --]

       Hello,

Building an arm-elf cross-debugging gdb 5.1.1 on OpenBSD breaks
because of the "unix vs. __unix__ issue". This, I note, has been
discussed before, but the fix ended up being NetBSD specific. Attached
are my changes.


Best wishes,
    --Daniel



[-- Attachment #2: ChangeLog --]
[-- Type: text/plain, Size: 208 bytes --]

2002-04-04  Daniel Néri  <dne@mayonnaise.net>

	* hostchan.h (__unix): Define when __unix__ (e.g. works on
	OpenBSD too).
	* unixcomm.c: Provide default serial/parallel device names for
	OpenBSD and NetBSD.

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #3: RDI patch --]
[-- Type: text/x-patch, Size: 1076 bytes --]

diff -u -r1.1 gdb/rdi-share/hostchan.h
--- gdb/rdi-share/hostchan.h    2002/04/03 01:32:26     1.1
+++ gdb/rdi-share/hostchan.h    2002/04/04 00:04:51
@@ -24,7 +24,7 @@
 #endif
 
 /* A temporary sop to older compilers */
-#if defined (__NetBSD__) || defined (unix)
+#if defined (__unix__) || defined (unix)
 #  ifndef __unix              /* (good for long-term portability?)  */
 #    define __unix    1
 #  endif
diff -u -r1.1 gdb/rdi-share/unixcomm.c
--- gdb/rdi-share/unixcomm.c    2002/04/03 21:59:45     1.1
+++ gdb/rdi-share/unixcomm.c    2002/04/04 00:02:39
@@ -94,6 +94,22 @@
 #define SERPORT2   "/dev/ttyS1"
 #define PARPORT1   "/dev/par0"
 #define PARPORT2   "/dev/par1"
+#endif
+
+#ifdef __OpenBSD__
+#define SERIAL_PREFIX "/dev/cua"
+#define SERPORT1   "/dev/cua00"
+#define SERPORT2   "/dev/cua01"
+#define PARPORT1   "/dev/lpt0"
+#define PARPORT2   "/dev/lpt1"
+#endif
+
+#ifdef __NetBSD__
+#define SERIAL_PREFIX "/dev/dty"
+#define SERPORT1   "/dev/dty00"
+#define SERPORT2   "/dev/dty01"
+#define PARPORT1   "/dev/lpt0"
+#define PARPORT2   "/dev/lpt1"
 #endif

[-- Attachment #4: Type: text/plain, Size: 36 bytes --]


-- 
Daniel Neri
dne@mayonnaise.net

Re: RDI on OpenBSD

[Richard Earnshaw]

 /* A temporary sop to older compilers */
-#if defined (__NetBSD__) || defined (unix)
+#if defined (__unix__) || defined (unix)
 #  ifndef __unix              /* (good for long-term portability?)  */

This is wrong.  NetBSD doesn't define "__unix__" or "unix".  Hence the 
original test.

R.

Re: RDI on OpenBSD

[Richard Earnshaw]

>  /* A temporary sop to older compilers */
> -#if defined (__NetBSD__) || defined (unix)
> +#if defined (__unix__) || defined (unix)
>  #  ifndef __unix              /* (good for long-term portability?)  */
> 
> This is wrong.  NetBSD doesn't define "__unix__" or "unix".  Hence the 
> original test.

Ok, forget that.  The compiler defines "unix" (and hence, "__unix__"), so 
that isn't a problem.

R.

Re: RDI on OpenBSD

[Richard Earnshaw]

> >  /* A temporary sop to older compilers */
> > -#if defined (__NetBSD__) || defined (unix)
> > +#if defined (__unix__) || defined (unix)
> >  #  ifndef __unix              /* (good for long-term portability?)  */
> > 
> > This is wrong.  NetBSD doesn't define "__unix__" or "unix".  Hence the 
> > original test.
> 
> Ok, forget that.  The compiler defines "unix" (and hence, "__unix__"), so 
> that isn't a problem.

Hmm, this is getting silly :-)

It turns out that I was more right the first time, than the second.

Some NetBSD platforms (in particular, those that are ELF) do not define 
either "unix" or "__unix__".  So we really do need that test for 
__NetBSD__ there.

R.

Re: RDI on OpenBSD

[Daniel Néri]

Richard Earnshaw <rearnsha@arm.com> writes:

> So we really do need that test for __NetBSD__ there.

Yep, it seems you're right. Anyway, I'll be happy if either __unix__
or __OpenBSD__ gets added to the condition in question.


Regards,
   --Daniel


-- 
Daniel Neri
dne@mayonnaise.net