cscope

Hi Darryl,

> [ This list has been pretty dead, and so I thought I'd throw out a
>   tidbit.  ]
Yeah sorry about this. Lots of things happening. As you know, SCO and
Caldera have merged (if that is the right word).

Thanks for doing the comparison between flex and lex. As for open
sourcing lex... may be a high possibility (who knows).

In general folks should start applying their work. I have several
patches/fixes to put in, may be a couple of days before I get the
chance, but lets get cscope 15.0bl3 out on the road. bl3 will become
15.0 final. The subsequent next version will be 15.1 or 16.0 depending
on the scope of enhancements.

Petr
> 
> FYI,
> 
>      I recently compiled cscope using flex, and I noticed that it seemed
> to run pretty slowly (indexing files, that is).  Well, I compiled cscope
> using different flex/lex options, ran some tests, and the results are
> pretty interesting.  Bottom line: the way cscope uses it, flex is a
> dog.  Old AT&T lex gives *MUCH* better results.
> 
>      Here are the results, run on a large-ish (~1000 files),
> multi-directory project:
> 
>                                 User+Sys (seconds)
> lex (AT&T)                       40.25
> flex -l (default)               247.68  (5.17X slower)
> flex (no -l)                    208.03  (6.15X slower)
> 
> Cscope was compiled using -g, and the system used was an old, somewhat
> slow, HP-UX box.
> 
>      Perhaps cscope should include an already-translated scanner.c?
> 
> [ Or, perhaps, SCO should also open-source lex? ;-) ]
> 
> --
>         Darryl Okahata
>         darrylo@...
> 
> DISCLAIMER: this message is the author's personal opinion and does not
> constitute the support, opinion, or policy of Agilent Technologies, or
> of the little green men that have been following him all day.
> 
> _______________________________________________
> Cscope-devel mailing list
> Cscope-devel@...
> http://lists.sourceforge.net/mailman/listinfo/cscope-devel

-- 
--------------------------------------------------------
Petr Sorfa                  Software Engineer
Santa Cruz Operation (SCO) soon Caldera 
430 Mountain Ave.           http://www.sco.com
Murray Hill 07974
NJ, USA
--------------------------------------------------------
Disclaimer: All my comments are my own and nobody else's
----------------------------------------------------------

On Tue Aug 01,2000 (06:16:24PM -0700), darrylo@... Okahata) wrote:
> [ This list has been pretty dead, and so I thought I'd throw out a
>   tidbit.  ]

Sorry - been a bit tied up with polishing an OS release...
 
> FYI,
> 
>      I recently compiled cscope using flex, and I noticed that it seemed
> to run pretty slowly (indexing files, that is).  Well, I compiled cscope
> using different flex/lex options, ran some tests, and the results are
> pretty interesting.  Bottom line: the way cscope uses it, flex is a
> dog.  Old AT&T lex gives *MUCH* better results.
> 
>      Here are the results, run on a large-ish (~1000 files),
> multi-directory project:
> 
> 				User+Sys (seconds)
> lex (AT&T)			 40.25
> flex -l (default)		247.68	(5.17X slower)
> flex (no -l)			208.03  (6.15X slower)


interesting - conventional wisdom has it that flex is much faster than lex.
perhaps its due the scanner description being still somewhat 
lex optimised/oriented rather than flex directed
 
any flex-head feel up to looking at the scanner description ??

> Cscope was compiled using -g, and the system used was an old, somewhat
> slow, HP-UX box.
> 
>      Perhaps cscope should include an already-translated scanner.c?

probably - that'd be cool for folks without either lex or flex and
unable/unwilling to find/build them ...
 
> [ Or, perhaps, SCO should also open-source lex? ;-) ]

perhaps there'll be fewer hurdles to doing so from here on in ....

-- 
- hops

Everything disclaimed (including disclaimer)
------<hops@...
Mike Hopkirk (hops)  |       Whenever possible steal code.
SCO Inc              |          Tom Duff. (ex) Bell Labs

Yeah... I've been busy at a client site for the last few months.
I'll make time to stop being lame soon...

* Mike Hopkirk(hops) <hops@...> [000802 14:21]:
> interesting - conventional wisdom has it that flex is much faster than lex.
> perhaps its due the scanner description being still somewhat 
> lex optimised/oriented rather than flex directed

I've heard that bit of CW as well... so I'm just as surprised.
 
> probably - that'd be cool for folks without either lex or flex and
> unable/unwilling to find/build them ...

Hmm... maybe we can have it as a "pre.scanner.c" file that depending
on ./configure options you either use it or whatever your machine has.
The problem is that lex and flex both use shared libraries... be it
libl or libfl... so someone will have to test how well this works
in practice.  Even then, the issue could concievably be poor library
procedures, etc.

> > [ Or, perhaps, SCO should also open-source lex? ;-) ]
> 
> perhaps there'll be fewer hurdles to doing so from here on in ....

That would be interesting as well. :)  I hope you guys at SCO are ok...

-- 
Joshua Uziel, Senior Linux Consultant, Linuxcare, Inc.
415.354.4878 tel, 415.701.7457 fax
uzi@..., http://www.linuxcare.com/
Linuxcare. Support for the revolution.

Joshua Uziel <uzi@...> wrote:

> Hmm... maybe we can have it as a "pre.scanner.c" file that depending
> on ./configure options you either use it or whatever your machine has.
> The problem is that lex and flex both use shared libraries... be it
> libl or libfl... so someone will have to test how well this works
> in practice.  Even then, the issue could concievably be poor library
> procedures, etc.

     Not a problem.  Any pre-generated file could have added routines
such that linking against any special library is not needed.  For lex, I
think that only yywrap() needs to be defined.  Besides, I don't know of
any platform that has libl or libfl as a shared library (this goes for
Linux, FreeBSD, Solaris, HP-UX, and AIX).

--
	Darryl Okahata
	darrylo@...

DISCLAIMER: this message is the author's personal opinion and does not
constitute the support, opinion, or policy of Agilent Technologies, or
of the little green men that have been following him all day.

* Darryl Okahata <darrylo@...> [000802 16:11]:
>      Not a problem.  Any pre-generated file could have added routines
> such that linking against any special library is not needed.  For lex, I
> think that only yywrap() needs to be defined.  Besides, I don't know of
> any platform that has libl or libfl as a shared library (this goes for
> Linux, FreeBSD, Solaris, HP-UX, and AIX).

Oh duh... you're right... and a good point.

Ok, so I guess then if this is decided as "a good thing", then someone
should check into CVS a "good" scanner.c ... as something other than
scanner.c ... pre.scanner.c or whatever.  Then a ./configure option
(like --no-flex) should be picked and implemented.

If needed, I can do the last part... I think. :)

-- 
Joshua Uziel, Senior Linux Consultant, Linuxcare, Inc.
415.354.4878 tel, 415.701.7457 fax
uzi@..., http://www.linuxcare.com/
Linuxcare. Support for the revolution.