Re: CVS commit: src/usr.bin/tic

[Roy Marples <roy%marples.name@localhost>]

On 04/05/2017 23:51, Robert Elz wrote:
>     Date:        Thu, 4 May 2017 14:07:33 +0000
>     From:        "Roy Marples" <roy%netbsd.org@localhost>
>     Message-ID:  <20170504140733.B327DFBE4%cvs.NetBSD.org@localhost>
> 
>   | Module Name:	src
>   | Committed By:	roy
>   | Date:		Thu May  4 14:07:33 UTC 2017
>   | 
>   | Modified Files:
>   | 	src/usr.bin/tic: tic.c
>   | 
>   | Log Message:
>   | Remove __VALGRIND__ guard .... just clean up.
> 
> 
> This breaks building on systems that have hcreate() but not hdestroy1()
> (like say, NetBSD 6).   tic is (apparently) required as a host tool, the
> build fails now as ...
> 
> #      link  tic/tic
> cc -O -I/usr/obj/testing/tools/include/compat -I/release/testing/src/tools/compat -DHAVE_NBTOOL_CONFIG_H=1 -D_FILE_OFFSET_BITS=64 -I/release/testing/src/tools/tic/../../usr.bin/tic/../../lib/libterminfo   -o tic tic.lo compile.lo hash.lo -L/usr/obj/testing/tools/lib -lnbcompat -lrt -lz
> tic.lo: In function `main':
> tic.c:(.text+0xf5b): undefined reference to `hdestroy1'

Yes, just received the same error from Andreas about this on a Linux host.

> Please revert this change, there doesn't seem to be a point to it
> (it is certainly just a waste of time to race around freeing memory
> and then immediately call exit() ... unless you are debugging memory
> leaks, say with valgrind).

Well, that's an interesting perspective.
As we know, parts of NetBSD code are used outside of NetBSD.
Should we exlcude RT OS's such as RTEMS from using our code in the
future by not doing this? Maybe a more common guard could be used?

Roy