Skip to contentSkip to conversation

quality control

DISCUSSION Hans Reiser 7 participants 10 messages Feb 6, 2006 – Feb 8, 2006 2 messages not in the archive

lore
Share & tools
lore
#2 lore raw
In reply to a message not in the archive

Dearest HCH,

Please consider adhering to a quality control process. No patch comes out of Namesys without a second person testing it (that includes compiling it). Users should not be burdened with code that has not been tested by a second person. Everyone makes mistakes of this kind, the difference is that some persons use a quality control process to avoid burdening more than one other person with them.

Hans

Quoted 55 lines · Jeff Mahoney
Quoted 34 lines · Hans Reiser

http://bugzilla.kernel.org/show_bug.cgi?id=6016

Summary: reiserfs doesn't build with REISERFS_FS_POSIX_ACL=n
Kernel Version: v2.6.16-rc2-g5b7b644
Status: NEW
Severity: normal
Owner: [email protected]
Submitter: [email protected]

Most recent kernel where this bug did not occur: 2.6.16-rc1
Distribution: Gentoo
Hardware Environment: amd64
Problem Description:

reiserfs doesn't build with REISERFS_FS_POSIX_ACL=n

Steps to reproduce:

$ make
 CHK     include/linux/version.h
 SPLIT   include/linux/autoconf.h -> include/config/*
 CHK     include/linux/compile.h
 CHK     usr/initramfs_list
 GZIP    kernel/config_data.gz
 IKCFG   kernel/config_data.h
 CC      kernel/configs.o
 LD      kernel/built-in.o
 CC      fs/reiserfs/xattr.o
fs/reiserfs/xattr.c: In function `reiserfs_check_acl':
fs/reiserfs/xattr.c:1330: error: called object is not a function
make[2]: *** [fs/reiserfs/xattr.o] Error 1
make[1]: *** [fs/reiserfs] Error 2
make: *** [fs] Error 2
Quoted 5 lines

Reverting ec191574b9c3cb7bfb95e4f803b63f7c8dc52690
---
[PATCH] reiserfs: use generic_permission

Use the generic_permission code with a proper wrapper and callback

instead

Quoted 3 lines

of having a local copy.
---
fixes the problem, but causes some warnings about unused symbols.

This was a patch from hch, not me. There's already a patch in -mm to fix it.

-Jeff

--
Jeff Mahoney
SUSE Labs

#3 lore raw
Quoted 11 lines · Hans Reiser
Quoted 8 lines · Jeff Mahoney
Quoted 4 lines · Hans Reiser

http://bugzilla.kernel.org/show_bug.cgi?id=6016
Summary: reiserfs doesn't build with
REISERFS_FS_POSIX_ACL=n
Kernel Version: v2.6.16-rc2-g5b7b644

This was a patch from hch, not me. There's already a patch in -mm to fix it.

Please consider adhering to a quality control process.

It's a GIT version of an RC patch for grief's sake! You don't seriously expect people to quadruple-check every trivial patch that goes into Linus GIT tree before sending it, do you? The whole point of the RC is to indicate that only smaller patches should be applied (and this one was for the most part) so that we can do some kind of global-kernel QC.

Quoted 1 line

Everyone makes mistakes of this kind, the difference is that some persons use a quality control process to avoid burdening more than one other person with them.

Precisely! The guy who reported the bug is the one person who was burdened with it. It will get fixed in the GIT tree, and only somebody who happened to test the dev tree between the two patches with that particular .config will have noticed.

BTW: Nice way to CC a private thread to a public list without the consent of all parties. You also made the grievous errors of (A) top-posting, (B) fullquoting without snipping irrelevant material, and (C) sending flamebait to the list (which I am so stupidly responding to).

Cheers,
Kyle Moffett

--
Debugging is twice as hard as writing the code in the first place. Therefore, if you write the code as cleverly as possible, you are, by definition, not smart enough to debug it.
-- Brian Kernighan

#4 lore raw
Quoted 11 lines · Kyle Moffett
Quoted 10 lines · Hans Reiser
Quoted 8 lines · Jeff Mahoney
Quoted 4 lines · Hans Reiser

http://bugzilla.kernel.org/show_bug.cgi?id=6016
Summary: reiserfs doesn't build with
REISERFS_FS_POSIX_ACL=n
Kernel Version: v2.6.16-rc2-g5b7b644

This was a patch from hch, not me. There's already a patch in -mm to fix it.

Quoted 8 lines
Quoted 1 line

Please consider adhering to a quality control process.

It's a GIT version of an RC patch for grief's sake! You don't seriously expect people to quadruple-check every trivial patch that goes into Linus GIT tree before sending it, do you? The whole point of the RC is to indicate that only smaller patches should be applied (and this one was for the most part) so that we can do some kind of global-kernel QC.

Eh, but you are expected to do that. If everybody sent in half-assed not tested patches "just because" it's a pre-rc, things would look bad. Compile testing is the least time consuming and easiest thing to to test, so you should at least do that. If nobody did that, no one would get snapshots tested because they would never compile for anyone.

For developers it's equally annoying. I typically update my tree every morning and run with that for the various stuff I'm working on, and it is really annoying to have to spent time on hunting down and fixing compile errors.

This mail is about the mentality displayed, not the specific change that spawned it.

Signature · 1 line
Jens Axboe
#5 lore raw
In reply to Kyle Moffett
Quoted 5 lines

BTW: Nice way to CC a private thread to a public list without the consent of all parties. You also made the grievous errors of (A) top- posting, (B) fullquoting without snipping irrelevant material, and
(C) sending flamebait to the list (which I am so stupidly responding to).

... which is a nice co-incidence in a thread on quality control :)))

Martin

#6 lore raw
In reply to Kyle Moffett
Quoted 4 lines · Kyle Moffett

It's a GIT version of an RC patch for grief's sake! You don't seriously expect people to quadruple-check every trivial patch that goes into Linus GIT tree before sending it, do you?

No quadruple check, but every patch going to Linus should get at least some basic testing and it's definitely suppose to compile at least in one .config combination.

Quoted 4 lines

The whole point
of the RC is to indicate that only smaller patches should be applied (and this one was for the most part) so that we can do some kind of global-kernel QC.

global kernel QA doesn't replace individual patch QA. Black box testing is no replacement for targetted tests.

-Andi

#7 lore raw
Quoted 4 lines · Andi Kleen
Quoted 1 line · Kyle Moffett

It's a GIT version of an RC patch for grief's sake! You don't seriously expect people to quadruple-check every trivial patch that goes into Linus GIT tree before sending it, do you?

No quadruple check, but every patch going to Linus should get at least some basic testing and it's definitely suppose to compile at least in one .config combination.

Well, yes, and it did. The problem was that if you turned off ACLs, it didn't work; only one or two variants of about 6 or 8 ways to configure reiserfs stopped working. Given that, I can't see how Hans is complaining about lack of QC. Nobody is going to test patches against every possible kernel configuration; that's why we do an RC, so that we can get a lot of different configs tested.

Cheers,
Kyle Moffett

--
I didn't say it would work as a defense, just that they can spin that out for years in court if it came to it.
-- Rob Landley

#8 lore raw
Quoted 6 lines · Kyle Moffett
Quoted 5 lines · Andi Kleen
Quoted 3 lines · Kyle Moffett

It's a GIT version of an RC patch for grief's sake! You don't seriously expect people to quadruple-check every trivial patch that goes into Linus GIT tree before sending it, do you?

Quoted 7 lines
Quoted 3 lines

No quadruple check, but every patch going to Linus should get at least some basic testing and it's definitely suppose to compile at least in one .config combination.

Well, yes, and it did. The problem was that if you turned off ACLs, it didn't work; only one or two variants of about 6 or 8 ways to configure reiserfs stopped working. Given that, I can't see how Hans

Look, it's really simple: lets say I make a change that has to do with PM, you do a quick compile test with and _without_ PM just to check you didn't screw anything up with that change. You change reiserfs acl stuff, you do a quick compile test with and without that configured.

It's a pretty standard procedure, and contrary to what you think, it _is_ required before submitting a patch. No one is asking anyone to check all possible configure options, but the interesting data set is typically extremely easy to guess looking at a change.

Signature · 1 line
Jens Axboe
#11 lore raw
In reply to Andi Kleen
Quoted 9 lines · Andi Kleen
Quoted 4 lines · Kyle Moffett

It's a GIT version of an RC patch for grief's sake! You don't seriously expect people to quadruple-check every trivial patch that goes into Linus GIT tree before sending it, do you?

No quadruple check, but every patch going to Linus should get at least some basic testing and it's definitely suppose to compile at least in one .config combination.

Right. We have now git-bisect, and it helped me to nail down a few bugs. Just now I track down some scsi or whatever breakage in -rc1. And guess what, not a single compile error so far, with a full featured config! So you guys better send tested patches, via akpm, to keep Linus tree in a reasonable shape.

Signature · 2 lines
short story of a lazy sysadmin:
 alias appserv=wotan
#9 lore raw
In reply to Jens Axboe
Quoted 10 lines · Jens Axboe

Look, it's really simple: lets say I make a change that has to do with PM, you do a quick compile test with and _without_ PM just to check you didn't screw anything up with that change. You change reiserfs acl stuff, you do a quick compile test with and without that configured.

It's a pretty standard procedure, and contrary to what you think, it _is_ required before submitting a patch. No one is asking anyone to check all possible configure options, but the interesting data set is typically extremely easy to guess looking at a change.

<rofl>

bix:/usr/src/op> find patches -name '*build-fix*' | wc -l

    533

bix:/usr/src/op> find patches -name '*fix.patch' | wc -l
5109

A lot of people don't make the slightest effort. But it's not a big problem, really. Silly build errors are reported early and are almost always trivial to fix. The major drawback is that they can wreck a -mm release for many testers.

#10 lore raw
Quoted 17 lines · Andrew Morton
Quoted 10 lines · Jens Axboe

Look, it's really simple: lets say I make a change that has to do with PM, you do a quick compile test with and _without_ PM just to check you didn't screw anything up with that change. You change reiserfs acl stuff, you do a quick compile test with and without that configured.

It's a pretty standard procedure, and contrary to what you think, it _is_ required before submitting a patch. No one is asking anyone to check all possible configure options, but the interesting data set is typically extremely easy to guess looking at a change.

<rofl>

bix:/usr/src/op> find patches -name '*build-fix*' | wc -l

    533
Quoted 7 lines

bix:/usr/src/op> find patches -name '*fix.patch' | wc -l
5109

A lot of people don't make the slightest effort. But it's not a big problem, really. Silly build errors are reported early and are almost always trivial to fix. The major drawback is that they can wreck a -mm release for many testers.

That's precisely the problem, it may be really simple to fix but often will stop people from testing.

Your fix count probably isn't totally accurate either, I bet a lot of these are fixups due to conflicts with other patches. I'm talking about the fact that someone sends Linus a patch which doesn't compile for the case you could (and should) have trivially checked. A little edumacation never hurt :-)

Signature · 1 line
Jens Axboe