Re: breaking -HEAD

On Tue, Feb 12, 2002, Henrik Nordstrom wrote:
> I am not very happy with intentionally breaking features HEAD.
>
> Should also note that there is a huge backlog of mostly tested
> features waiting to get committed to HEAD.

I'm planning on working through them in some reasonably sane
fashion (starting with andre's memory allocator) after I get this
code committed.

I don't want to break squid-HEAD either but I think the changes
I'm doing warrant some temporary breakage. The API itself isn't going
to change for a while - I'm going to concentrate now on doing some
of the stuff that the new API will let us do and get it all stable
before I work on the next bit of API breakage. Chances are the next
bit of API breakage is going to involve splitting out storeGet()
et al into a callback routine - which again will break the cache digest
code.

I will go back and get the digests code going well before squid-HEAD
becomes stable.

Henrik - I think that the commloops stuff is the beginning of a very,
very large shift in the squid internals. My plan is to get this and
the other pending projects into HEAD and tested thoroughly before the
pending work becomes too stale. If I wasn't worried about this I'd just
keep going until I've completely redone the storage manager and network
code - then noone's work will be committed since all the branches will
be so far apart.

Once everything thats pending is in HEAD and the bugs are ironed out
I'll move forward to the next round. That way, all the new nifty stuff
that people have been working on goes into -HEAD and their combined
behaviour gets dealt with (and fixed if needed!).

If you are still against it then I'll keep going along with the digest
code until I get it finished - but I don't have any timeline for that
and I'd rather the work not rot because of one module.

Adrian
Received on Tue Feb 12 2002 - 18:44:09 MST