[Savannah-hackers] Re: submission of Bash with debugging support - savannah.gnu.org

[R. Bernstein]

Mathieu Roy writes:
 > It would be really better if your work become part of the GNU Bash
 > project.

I agree with you wholeheartedly. The problem is there has been a bit
of history which suggests this isn't going to happen.

 > 
 > Apparently, GNU Bash people use their savannah group only to manage
 > their web page.

If you look at the "bash" project in contrast to autoconf or emacs, or
glibc, you'll see there's only person for the entire project and no
developers. I don't think is by accident.

Although I applaud the work of Chet Ramey and think he's done a
marvelous job in supporting bash: fixing bugs, making improvements,
providing support to question posted in newsgroups and probably
privately as well. However has run bash as a largely one-person
project.

About two years ago when I discovered SIGDEBUG was added, I first got
intrigued with the possibility of having a debugger. I realized then
that a bit of work was needed inside bash to add support to make writing
a debugger possible. The two most critical things I mentioned in
comp.unix.shell on Date: 2000-09-17 07:59:06 PST.

See:
  
http://groups.google.com/groups?q=bashdb+Bernstein+group:comp.unix.shell&hl=en&lr=&ie=UTF-8&selm=wihn1h7t686.fsf%40panix3.panix.com&rnum=2

However this note was sent after contacting Chet and getting no
reply. I got the sense that there was not even interest in a debugger
at the time. I am sure I sent him the patches (or perhaps an earlier
version) that are mentioned in the posting. The fact that I had to put
the patches and debugger elsewhere were indicative of the fact that I
couldn't get them into the main-line code. There was no
discussion. This is very different from my experience with other GNU
projects. Either code is accepted or there is some *communication* and
discussion about issues. (The only communication I recall was lack of
interest.)

Anyway, because of the serious flaws and perceived lack of interest in
getting those flaws fixed, I dropped interest.  Recently I saw in bash
version 2.05b that both of the most serious problems I mentioned in
2.04.1 had been fixed. Of course, Chet didn't contact me; I didn't
even know he was considering changing either of those. Other
lesser-important issues still need to be addressed to make debugging
support possible.

If debugging in bash is to ever get done, something different about
the project management probably has to happen. Sadly, I understand
this is a people problem more than a technical problem.

 > You should contact this project admin to get in touch with the
 > developers. And so, you'll be able to work with them with the tools
 > they use.

Again you are correct and I will do. Frankly, I don't hold out much
hope though, and the second time around I'm not as patient. As for the
"tools" used, I don't see version control Id's lines on the files. I
have a feeling that the project would benefit from Savannah.

My own first choice for where the code should live would be Savannah,
of course. However bash is GPL, and so are my changes. There are other
sites that I'm sure would be willing to host this in an open
project-management style. My suggestion then is to start this as a
Savannah project. It could only help using Savannah for the main-line
bash project. Of course if the debugger part gets folded into the
main-line code, I'd be all the more happy to eliminate the project
then. 

Again, I appreciate any thoughts or help you might bring to bear on my
dilemma.