Notes on Code Crusader (Was: Re: NT recommend?)

Notes on Code Crusader (Was: Re: NT recommend?)

Post by John Lind » Fri, 15 May 1998 04:00:00



A friend pointed out this post on Deja News, but I haven't seen it on
usenet yet.  Since this probably just means that our news feed is slow,
I decided to post a reply now, to get it out of the way.

-----


in which he discusses Code Crusader:

Quote:> This package appears to be very immature (though, granted, it looks like
> it's only been worked on for six months).

That's the way it goes when one doesn't get paid to do it and works only in
one's spare time.  The only real biggie is the complete interface to make,
though, and the program "makemake", which is included in the Code Crusader
distribution, provides a plain text project interface.

Quote:> Some of the "features under development" and "planned features" are
> things which I really depend on.

I'm working as fast as I can :)

Quote:> Here are some of the things I _demand_ in a visual design environment:

> De* integrated into editor (this actually wouldn't be difficult to
> do in Linux, all you need to do is set up a pipe to talk to gdb).

Wrong!  We are actually doing this (Code Medic) and it's a ***.  gdb's
output is designed only for human readability (and only marginally at
that!), so one needs an AI to actually parse all the output.  Bison
certainly isn't up to it.  We tried.

Quote:> Extensibility via scripting language, or embedded modules

Then you want emacs, unless you want to write C++ code to extend Code
Crusader.

Quote:> Powerful code / class browser (to do this right, you need to parse the C
> code into some format that a computer can read, like a listing file.
> AFAIK, gcc cannot do this.)

Code Crusader does have a C++ class tree.

Quote:> Hypertextual (and consistent) documentation for all included libraries
> (for example, in Linux, it would be _very_ useful to have hypertext
> documentation for the Motif widget hierarchy which was as good as the MFC
> class hierarchy documentation (but all they give is man pages).

One can grumble when Sun or HP doesn't do this, but one can't expect
miracles from a volunteer effort like Linux, since everybody simply
contributes what they themselves need.

The GNU Info pages are hypertext.

Code Crusader provides a kind of hypertext within the man pages.

Another point to remember is that UNIX isn't a single OS.  The functions
aren't the same across platforms.  Sun's libraries are even different for
different versions of their UNIX.  (SunOS vs Solaris)  Since I want Code
Crusader to be useful on all versions of UNIX, I can't provide more than an
interface to the standard tools that come with the system.

Quote:> Interface builder (on Linux preferably Motif or Uil, not something weird).

That's up to the designers of each individual library.  I provide such a
system with my JX application framework, but it's not appropriate for Code
Crusader since each person has their own personal favorite:  JX, Qt, gtk,
Motif, etc, etc.  Personally, I would never touch Motif (obviously, since I
wrote JX) because it's not C++.

John

 
 
 

Notes on Code Crusader (Was: Re: NT recommend?)

Post by TERENCE MURP » Fri, 15 May 1998 04:00:00




>A friend pointed out this post on Deja News, but I haven't seen it on
>usenet yet.  Since this probably just means that our news feed is slow,
>I decided to post a reply now, to get it out of the way.

Hi,

Thanks for taking the time to respond to my message.  Please note that
I wasn't really trying to specifically attack your product, but rather
the general lack of a product comparabe to VC++ 5.0.

What I sort of hate to say here, but need to, is that all you really
posted was _causes_ of the some of the problems/limitations, due to
limitations (at some level) in the UNIX/Linux environment/philosophy.

For every point you made, except for the emacs and code browser, I can
respond by simply saying that the point is not a problem in commercial,
non-UNIX OS'es such as NT (i.e. depending on a college student to
develop code in his free time, needing to parse human readable de*
output, not being able to provide a GUI builder because there are too
many GUI's, and not being able to provide good library documentation
because the OS'es don't).

I'm sorry, but you have really strengthened my position here.  ;-)

Cheers,

Terry Murphy

 
 
 

Notes on Code Crusader (Was: Re: NT recommend?)

Post by Frank T. Lofar » Fri, 15 May 1998 04:00:00



>A friend pointed out this post on Deja News, but I haven't seen it on
>usenet yet.  Since this probably just means that our news feed is slow,
>I decided to post a reply now, to get it out of the way.
>-----

>in which he discusses Code Crusader:
>I'm working as fast as I can :)
>> Here are some of the things I _demand_ in a visual design environment:

>> De* integrated into editor (this actually wouldn't be difficult to
>> do in Linux, all you need to do is set up a pipe to talk to gdb).
>Wrong!  We are actually doing this (Code Medic) and it's a ***.  gdb's
>output is designed only for human readability (and only marginally at
>that!), so one needs an AI to actually parse all the output.  Bison
>certainly isn't up to it.  We tried.

ARGH! Don't make life so difficult for yourself! 1/2 ;)

You can get and change the gdb source.

Hack gdb so it has a machine-readable output format,
and have it use something like set format human or set format
machine. Just like GNU stty has the nice output, and
the machine-readable format:

d06:5:cbd:3b:3:1c:7f:15:4:0:1:0:11:13:1a:0:12:f:17:16:0:0:73

The gdb people should be willing to include such a hack.
Hacking gdb isn't so bad, I hacked gdb so I could use a set
command to change disassembly to assume 16 bit addresses
and/or words. (didn't submit it yet, argh)

Alternative: You could just use some of the gdb code directly.

Either way is easier, cleaner, and less prone to fail if you do it
than doing it the hard way.

Good luck.