Re: [Lxr-dev] Reaching 1.0

[Malcolm B. <ma...@br...>]

Hi,

Per Kristian Gjermshus wrote:
> > What do you think - are we at the point where we could declare a 1.0
> > release and move on to the next major development cycle, or are we >
> still
> > missing key features, bug fixes or stability?
> 
> There are currently 6 open bugs in the bugtracker. I don't think we
> should declare 1.0 before all of them are resolved. We could either
> decide that a bug is no problem for 1.0 or we should fix it.

I agree, we should go through all the bugs and either postpone them to
the next release or fix them for 1.0.

Currently there are the following bugs outstanding:

447980   Generic config missing
463138   Unable to enter full glimpse REs
469413   Non-symbols can be mistaken for symbols
471858   Some characters in files create trouble
476695   Java interfaces display as docs
476773   Shouldn't install global signal handlers
476775   mod_perl coding style should be checked
481573   requires non-free software for searching
481597   Should index X::Y() as well as Y()

Of these, I think "Generic config missing", "Non-symbols can be mistaken
for symbols", "Shouldn't install global signal handlers", "mod_perl
coding style should be checked" and "Should index X::Y as well as Y" can
all be futured to post 1.0.

That leave 4 bugs remaining.  I have a fix in progress for 476695 and I
can see a way to fix 481573.  471858 & 463138 are the same underlying
bug to do with how aggressively we wash incoming http parameters.  The
signal handlers fix is trivial to remove, but replacing with a good
solution is more difficult, so maybe we should future this one.

All in all, there's not too much left to do I feel.  Of course, there's
still the lurking "unable to locate module Foo" bug that seems to crop
up periodically, but without a reliable way to reproduce it nor a good
diagnoses I don't know what to do about it.

> Another thing is that I think that 1.0 should be able to index the
> entire linux-kernel and run on lxr.linux.no. It must of course be me who
> does this testing.

It would be very good to be able to showcase the new version on
lxr.linux.no.  Are you likely to be able to install the new version any
time soon?

It may also be possible to install a demo site on the sourceforge
webservers, possibly indexing some of the other projects on SF.  I
haven't yet investigated how easy this will be to set up.

> We should also decide on which database backends to support. We should
> not ship backends that do not work. Have anyone ever gotten the DBFile
> backend to work?

I know of no-one who has got it to work - unless someone steps up to
make it work and maintain it, I suggest it should be dropped from the
release.

There's also a PR job to do with all the various sites round the web
that run the LXR to try to convince them to upgrade to the new version. 
So far I know of:

cvs.gnome.org
lxr.mozilla.org

who are running the 0.3 codebase.  If anyone knows of any others, please
let me know so I can contact them and suggest they might want to try the
newer version.

Cheers,

Malcolm