On 13 Oct 1999 05:41:35 +0900, [EMAIL PROTECTED] wrote:

>The debug code
>say size;'cmd/c dir chkconfig.sed';
>checks if the arguments to stream() is all right, and it seems OK.
>The stream() line is setting the pointer to the end of the file

Ahh... This might be my idiosyncrasy, but with Rexx I've always
difficulties to understand what's going on in the code... :-(

>chkconfig.sed
>With procedural REXX (non ANSI)
>     rc=stream('chkconfig.sed','c','seek + 'size)
>is correct and
>     rc=stream('chkconfig.sed','c','seek + 'size' write')
>is wrong, while with ObjectREXX (whose procedural portion is ANSI)
>     rc=stream('chkconfig.sed','c','seek + 'size' write')
>is correct and
>     rc=stream('chkconfig.sed','c','seek + 'size)
>is wrong!

Ahemhh.

>The line
>if rxfuncquery(SysAddRexxMacro) then
>tries to figure out which version of REXX interpreter is loaded.

And it fails... Searching wildly in the Short Reference digged out

        PARSE VERSION (could this help? I doubt that, though.)

>Are we going to use cmd /c "copy A+B C" ?
>Where A is the first half of chkconfig.sed, B is the last half,
>and C is the script to feed sed. 

Yes. Too simple/another hidden bug? I've always done a simple 'copy/v a
c' and then a stdout redirection '>> c'. Recommend being careful to
limit the usage of the nowadays outdated old 16bit protected mode
standard utilities, if possible: They are not so well maintained by IBM
(aka buggy, 'cause they stem partly from M$ hacks), unexpected problems
arise especially with certain non-US versions of OS/2.

>(cat is not an option)

I agree, too many too broken ports floating around (, e.g. the one I've
been using from leo).

Greets,

        Arnd

PS: Ceterum censeo, Rexx is a headache ;-) 

But this is not an imminent problem, though. On the _very_ long run, I
think, the process of building and configuring LyX should not be forked
into several different scripting languages. Best thing would be a
uniform and consistent method, which should - most prominently - be
easily maintainable. Just think about the code maintainers doing a last
minute change: Now they cannot without possibly breaking the build and
reconfigure process. 

Reply via email to