[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Gnu-arch-users] changes: tree shows no revisions in version ?

From: John A Meinel
Subject: Re: [Gnu-arch-users] changes: tree shows no revisions in version ?
Date: Fri, 10 Dec 2004 08:53:33 -0600
User-agent: Mozilla Thunderbird 0.9 (Windows/20041103)

Andrei A. Voropaev wrote:

On Thu, Dec 09, 2004 at 10:23:36AM -0600, Gustavo Córdova Avila wrote:
Andrei A. Voropaev wrote:

Well. It took me quite a few tries before I've figured this out. If
assigning different name to archive is so dangerous, then why is it
allowed? And if it is allowed, then why later usage of alternative name
leads to problems?


Well, for one thing, you might know just what you're doing,
and tla is eminently flexible in that account, it lets you
shoot yourself in the foot --IF you don't know what you're
Me, I find tla's "do what you want, it's you're life" attitude
refreshing, I absolutely hate when a tool starts applying
artificial restrictions and treating me (it's user) as an
ignoramus; we all start as such, but we learn, and having a
restrictive tool takes away the delight of learning, because
then your tools don't fit quite as nicely as before, so,
IMO, tla "does the right thing".

Hm. Interesting interpretation. I always understood this princip
differently. Take any calculator. It is also a tool. This tool does not
forbid me to divide by zero. But it does not crash when I do it, so I
can continue to use it. In my situation, tla didn't forbid my collegue
to checkout a tree under wrong name. But after that it refused to work
with it, leaving all his changes in the air. That's an equivivalent of
crash. tla restricted him and me in our work.

What is so restricting in not allowing to checkout a tree that wouldn't
be usable anyway?

Wouldn't it sound better if my question was "Why tla doesn't allow a
checkout of my tree" and your answer: "Because you are doing the wrong
thing. You shouldn't checkout from wrong archive name. RTFM." :) And
that would be so good for us, because then we wouldn't have to waste our
time trying to figure how to save all the updates to the tree.

It's so pleasant to learn, when one discovers new and new features. And
it sucks, when one has to learn more and more gotchas.

All of the above was not said as a reproach to tla :) Just an
explanation why I think that this behaviour should be considered a bug
and not the feature :) And of course, no demands to fix it. After all
this is free tool :)


I would venture there are few reasons why this bug has not been fixed.

  1. Most of us have learned that it's much better to let tla
     auto-detect the archive name, so people like me who have used it
     for a while don't run into this problem very often. A bug unseen
     is one unfixed.
  2. I think it is around because early versions of tla could not
     autodetect the archive name. Also, as mentioned you do need to
     still sometimes register $archive-MIRROR, or $archive-SOURCE. I
     believe someone submitted a patch to allow you to give an option
     to register archive, rather than specifying the name (much better).
  3. Also, someone at least talked about a patch to disallow
     registering incorrect names. If they did submit one, the only
     reason it might not be in the current tla is that it got lost in
     the shuffle. We had a bug-tracker running, but then we had a lot
     of churn where some work got done, but in the end it wasn't
     accepted, so the bug-tracker got very out of sync with mainline. I
     believe this is also being worked on.

Anyway, I agree that tla should never let you register an archive with the wrong name. And it is a bug to let you do so. I will submit a bug report on arch-dev, so at least people are reminded about it.


Attachment: signature.asc
Description: OpenPGP digital signature

reply via email to

[Prev in Thread] Current Thread [Next in Thread]