Suggestions to help the tree...

karl11 · Guru Joined: 25 Jun 2002 Posts: 469 Location: Raleigh, NC

Suggestions to fix problems from monkeys and tree post that was locked....please be constructive if you reply....sure it was flamebait, but you people have no clue how hot my blood boils to see stupid mistakes like that being made when I'm used to seing high quality stuff.

1) All ebuilds must be held to scrutiny. That is, they should all be checked for syntatical errors by the person checking them in to cvs. Perhaps cvs isn't the best idea for the tree. Other alternatives should be discussed. Namely, before anyone adds a build to a tree, they should make sure syntax errors are cleared.

2) All ebuilds must compile on at least one or two boxes before hitting the official tree. I don't care if you have an ebuild masked. Masked should still at least compile, or there shouldn't be an ebuild in the tree. You've masked so much good stable things for so long that masking is just a way to keep people from using developmental software or untested software. It no longer means packages that won't compile. So make absolute sure that packages have compiled on at least one or two boxes before putting it in. If an ebuild is masked in package.mask, then it doesn't have to comply with this rule. But all ~ masked packages should compile.

3) Actively criticize those who put an ebuild in the tree that violates the above. People need to be slapped on the hand for rushing something into portage, especially for syntax errors or logical errors in the ebuild.

I'm not meaning to be actively against gentoo or the way things are done, but I do believe that this should not be a problem and that the problem rises from people trying to get things out the door too quickly instead of taking the time to do it right. Let's not be the folks who don't care about the end user.

Karl

Lovechild · Posted: Sun Aug 03, 2003 7:21 pm Post subject:

Finally someone but me who has seen the light - Gentoo is not perfect in every way and form...

I already once tried to reform the way ebuild QA was done - and it ended up in a messy flamewar with very little actual constructive work being done.

I think Gentoo needs a reformation badly since it's users are no longer regarded or used as source for improvement.

but then again... I could be wrong...

avenj

avenj · Posted: Sun Aug 03, 2003 7:26 pm Post subject:

avenj · Posted: Sun Aug 03, 2003 7:28 pm Post subject:

karl11 · Guru Joined: 25 Jun 2002 Posts: 469 Location: Raleigh, NC

My idea is to only treat people with the same respect that they treat me.

Kesereti · Guru Joined: 07 Nov 2002 Posts: 520

I may not be a moderator on the forums...but I can see where this is going...no more back and forth about who said what and under what conditions, etc, please =) Let's try and keep this a civil discussion about ideas for making Portage (and Gentoo!) better than it already is...the name-calling and whatnot can be done in PM's ^_^

avenj · Posted: Sun Aug 03, 2003 7:37 pm Post subject:

Lovechild · Posted: Sun Aug 03, 2003 7:41 pm Post subject:

avenj: pitting karl11 against Foser, the most userunfriendly developer Gentoo has had to date, just isn't fair - the man is a walking flamewar.

I think karl11 is a brilliant guy, I've had nothing but positive feedback from him in the past - a true asset to the Gentoo community. If it wasn't for him and Lin-Matt we would have no place for quality out of tree ebuilds. I myself have a few ebuilds on bmg and I've gotten a lot of good feedback on the quality of my job for the users - in the event I made a mistake, they do let me have it.. and that's okay because most of them provide a fix as well. actively and constructively criticizing developers is a good thing.

As for the point about users. When I first joined Gentoo I was amazed at the good mood at surrounded it's development, the users were not so much users as they were part of the experience that is Gentoo - their ideas were reviewed and judged - now it seems users are considered more of a problem than a help. We are an asset to be harnessed not a problem to be fixed.

karl11 · Guru Joined: 25 Jun 2002 Posts: 469 Location: Raleigh, NC

The foser thing is an over and done with deal, but goes even further than what you saw in that bug report. There were personal issues which were hashed out in pms, emails, and forum posts that you never saw. Until you've read all of the dialogue, you really aren't in a position to comment on my replies on that bug. Please read down through the whole bug and also read all the threads concerning gnome development ebuilds before saying anything else on that topic.

Ok...so I was wrong with my monkey/tree post. However, there are some things that should be jumped on and they are stupid mistakes like syntax errors. What business do you have putting something that doesn't even work into the tree? I gave my suggestions as a constructive criticism in the first post on this thread. Go back and read it.

I'm not saying you should shoot a developer dead on the spot for getting something wrong as you tend to think I'm going overboard here. If a syntax error exists in an ebuild, the developer responsible needs to be notified and asked to try and not make the same mistake again or not rush things out the door so fast. This tends to be a problem around here.

Karl

avenj · Posted: Sun Aug 03, 2003 7:53 pm Post subject:

avenj · Posted: Sun Aug 03, 2003 7:56 pm Post subject:

karl11 · Guru Joined: 25 Jun 2002 Posts: 469 Location: Raleigh, NC

If you believe that your life will become all that more difficult for someone pointing out what you did wrong, then maybe it's time to re-evaluate your views on life, drop the holier than thou attitude, and let people tell you where you went wrong and then learn from your mistakes. Sample correspondence between me and a developer who has made a mistake:

-----------------------
dev,
I would like to point out that I believe I've found a mistake in your work. I might not be correct in my finding, but it bears looking at. In line 15 in your ebuild foo-1.2.3.ebuild, you are missing a " at the end of the statement. If you would, please fix this (if it is indeed an error) and try not to commit the same type of mistake to the tree in the future. Gentoolkit has some great debugging tools to help you out in that. Thanks for your contributions to gentoo and I hope this bug report has been a help to you!

Karl
---------------------

If that makes your life harder, then take a 3 week vacation.

Karl

p.s. As lovechild said, over at breakmygentoo, we do criticize people for their mistakes, but in the same regard, we also help them out and show them why they are mistakes. Keep in mind, this whole conversation started over the fact that the portage tree is getting largely out of control.

avenj · Posted: Sun Aug 03, 2003 8:21 pm Post subject:

karl11 · Guru Joined: 25 Jun 2002 Posts: 469 Location: Raleigh, NC

avenj -- I think we simply had a miscommunication that for the most part I'd like to say has been cleared up based on our last two posts. Perhaps in the future I will present things in a more plausible way to everyone who comes to the forums. Thanks for taking the time to actively discuss the issues though. I do appreciate your constructive comments.

Now back to the subject at hand based on the subject of this thread.....Suggestions for how to better maintain the tree folks? I do believe that each ebuild needs to be tested for syntax errors upon entry to the cvs tree.

Karl

aethyr · Veteran Joined: 06 Apr 2003 Posts: 1085 Location: NYC

karl11: what kind of syntax errors are we talking about?

If they're just bash script errors, I imagine there's some program that can check the validity of bash scripts out there already.

If they're errors that require actual thought to resolve, that might be more difficult.

lisa · Posted: Sun Aug 03, 2003 8:53 pm Post subject:

avenj · Posted: Sun Aug 03, 2003 8:57 pm Post subject:

Carlo · Developer Joined: 12 Aug 2002 Posts: 3356

avenj · Posted: Sun Aug 03, 2003 9:01 pm Post subject:

Pythonhead · Posted: Sun Aug 03, 2003 9:02 pm Post subject:

Did you consider that he did have a working ebuild and made an error and submitted the wrong one?

What good is any proposed system if people can still make human errors and submit the wrong ebuild?

avenj

karl11 · Guru Joined: 25 Jun 2002 Posts: 469 Location: Raleigh, NC

aux_get(): (2) Error in net-misc/ksambaplugin-0.4.2 ebuild.
Check for syntax error or corruption in the ebuild. (--debug)

This one is still outstanding...the other bug has been fixed but was simply missing a "

It had simply not been tested before being committed.

As to writing a utility to verify syntax....hmmm....we'll have to see...I might write such a thing for PortageSQL 0.3. 0.2 is coming out sometime here in the near future....

Karl

handsomepete · Posted: Sun Aug 03, 2003 9:23 pm Post subject:

Here's an idea; no idea how difficult it would be to implement in the real world:

Have something that watches for (or gets some sort of notification of) updates to the portage tree. Then, when that notification in whatever form is received, emerge it. If the emerge exits badly, send an email to a contact address saying that a test of x emerge failed on y hardware at z time.

I have a k6-2 I'm willing to donate to the cause that isn't doing anything - it'd only take a couple people. Granted, it's an ugly stopgap (but real world testing) measure... but just a thought. *shrug*

Carlo · Developer Joined: 12 Aug 2002 Posts: 3356