Skip to contentSkip to conversation

2.5.60 cheerleading...

DISCUSSION Jeff Garzik 14 participants 37 messages Feb 12, 2003 – Feb 20, 2003

lore
Share & tools
lore
#0 lore raw

Just to counteract all the 2.5.60 bug reports...

After the akpm wave of compile fixes, I booted 2.5.60-BK on my Wal-Mart PC [via epia], and ran LTP on it, while also stressing it using fsx-linux in another window. The LTP run showed a few minor failures, but overall 2.5.60-BK is surviving just fine, and with no corruption.

So, it's working great for me :)

#1 lore raw
Quoted 6 lines · Jeff Garzik

Just to counteract all the 2.5.60 bug reports...

After the akpm wave of compile fixes, I booted 2.5.60-BK on my Wal-Mart PC [via epia], and ran LTP on it, while also stressing it using fsx-linux in another window. The LTP run showed a few minor failures, but overall 2.5.60-BK is surviving just fine, and with no corruption.

Can you send me your list of failures and version of LTP? I'd like to make sure they match up with the known list of problems.

Thanks,
Paul Larson

#2 lore raw
Quoted 8 lines · Paul Larson
Quoted 3 lines · Jeff Garzik

Just to counteract all the 2.5.60 bug reports...

After the akpm wave of compile fixes, I booted 2.5.60-BK on my Wal-Mart PC [via epia], and ran LTP on it, while also stressing it using fsx-linux in another window. The LTP run showed a few minor failures, but overall 2.5.60-BK is surviving just fine, and with no corruption.

Can you send me your list of failures and version of LTP? I'd like to make sure they match up with the known list of problems.

Version is CVS-latest, checked out last night.

Some of the failures were obvious: kernel module syscalls were failing, as I would expect them to if they had not been updated for 2.5.x module support. There were also a couple signal-related things, IIRC. Unexpected failures included some file locking test failures.

Anyway, don't worry... you will be getting more concrete info soon :) I am tracking down right now why the 2.5.x floppy driver ends all I/O requests with an error. After that, I'll have output from a full LTP run for you :)

#3 lore raw
In reply to Jeff Garzik
Quoted 1 line · Jeff Garzik

Just to counteract all the 2.5.60 bug reports...

/me tries to blank vision of Jeff in cheerleader outfit.

8-)

Dave

Signature · 2 lines
| Dave Jones.        http://www.codemonkey.org.uk
| SuSE Labs
#4 lore raw
Quoted 5 lines
Quoted 1 line

Just to counteract all the 2.5.60 bug reports...

/me tries to blank vision of Jeff in cheerleader outfit.

8-)

That reminds me, I used this as a placeholder on my website for a while, why don't we use it as an oops graphic!?

root@darkstar:~# cat cheerleader
      |\/|           |\/|
     /****\ /|||||\ /****\
     \****/// ^.^ \\\****/
      |/\\//|  V  |\\//\|
         \\|\_____/|//
        /|\\|     |//|\
        |||\\|   |//|||
        ||||\     /||||
             |   |
            /     \
           /       \
           WWWWWWWWW
            | | | |
            | | | |
            | | | |
            | | | |
           / /   \ \
          /_/     \_\

John.

#5 lore raw
In reply to Jeff Garzik

Jeff Garzik wrote

Quoted 1 line

So, it's working great for me :)

I always wonder what should be said by testers when beta software actually *works*. ;)

I've been successfully running 2.5.60 on my Tycho system since a few minutes after Linus announced it. The major bugs I'd had in 2.5.59 were with Alsa, which seems to be running well in 2.5.60.

Tycho
Debian 'sid' GNU/Linux, Kernel 2.5.60 SMP (custom) 2.8GHz Pentium 4 w/HT, Intel D850EMV2, 533MHz FSB 512 MB PC800 RDRAM, 80GB Maxtor D740X 7200 RPM ATA-100 HD WinFast A170, GeForce 4 MX 440 GPU, 64MB DDR

I wonder if testers should post configs that worked, so developers can compare against configs that failed?

--
Scott Robert Ladd
Coyote Gulch Productions (http://www.coyotegulch.com)
Professional programming for science and engineering; Interesting and unusual bits of very free code.

#6 lore raw
Quoted 11 lines
Quoted 1 line

So, it's working great for me :)

I always wonder what should be said by testers when beta software actually *works*. ;)

I've been successfully running 2.5.60 on my Tycho system since a few minutes after Linus announced it. The major bugs I'd had in 2.5.59 were with Alsa, which seems to be running well in 2.5.60.

I wonder if testers should post configs that worked, so developers can compare against configs that failed?

Feel free to open a 'bug report' on my bug database just to upload a working .config - I.E. you can use it as a somewhat-organised, and comment-onable .config storage repository, if that's any use.

http://grabjohn.com/kernelbugdatabase/

If there is sufficient interest, I can add some code to pick out the common config options out of a set of config files, or whatever you want. Just ask, and the feature will be, (eventually), added :-)

John.

#7 lore raw
In reply to Scott Robert Ladd
Quoted 2 lines · Scott Robert Ladd

I wonder if testers should post configs that worked, so developers can compare against configs that failed?

Please don't. .configs tend to be huge, and a flood of 'me too' type posts will increase size of linux-kernel dramatically.

Dave

Signature · 2 lines
| Dave Jones.        http://www.codemonkey.org.uk
| SuSE Labs
#8 lore raw
In reply to Jeff Garzik

This brings up an interesting point. It seems like it's very common to have a release that doesn't boot, or produces immediately obvious problems. I'm curious if you do any testing (LTP or otherwise) on the kernels you intend to release. If not, would it be possible for someone to do this? I know we could never catch every problem, but at least the annoying, immediately noticeable problems could be caught and fixed quickly.

If you wanted to do this, I think that would be great. If you don't have time, I understand but would you be ok with me or anyone else doing at least a quick sniff test before release? It doesn't have to be anything fancy or time consuming. I'm not looking to add delays, just a small amount of extra testing before release so that hopefully more people will be willing to try the releases on their systems.

Thanks,
Paul Larson

Quoted 15 lines · Jeff Garzik

Just to counteract all the 2.5.60 bug reports...

After the akpm wave of compile fixes, I booted 2.5.60-BK on my Wal-Mart PC [via epia], and ran LTP on it, while also stressing it using fsx-linux in another window. The LTP run showed a few minor failures, but overall 2.5.60-BK is surviving just fine, and with no corruption.

So, it's working great for me :)

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

#9 lore raw
Quoted 4 lines · Paul Larson

This brings up an interesting point. It seems like it's very common to have a release that doesn't boot, or produces immediately obvious problems. I'm curious if you do any testing (LTP or otherwise) on the kernels you intend to release.

In the words of our fearless leader:

"regression testing"? What's that? If it compiles, it is good, if it boots up it is perfect.

:-)

Quoted 33 lines

If not, would it be possible for someone
to do this? I know we could never catch every problem, but at least the annoying, immediately noticeable problems could be caught and fixed quickly.

If you wanted to do this, I think that would be great. If you don't have time, I understand but would you be ok with me or anyone else doing at least a quick sniff test before release? It doesn't have to be anything fancy or time consuming. I'm not looking to add delays, just a small amount of extra testing before release so that hopefully more people will be willing to try the releases on their systems.

Thanks,
Paul Larson

Quoted 15 lines · Jeff Garzik

Just to counteract all the 2.5.60 bug reports...

After the akpm wave of compile fixes, I booted 2.5.60-BK on my Wal-Mart PC [via epia], and ran LTP on it, while also stressing it using fsx-linux in another window. The LTP run showed a few minor failures, but overall 2.5.60-BK is surviving just fine, and with no corruption.

So, it's working great for me :)

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

Signature · 2 lines
Grief can take care of itself, but to get the full value of a joy you must
have somebody to divide it with. -- Mark Twain
#10 lore raw
Quoted 10 lines · Edesio Costa e Silva
Quoted 4 lines · Paul Larson

This brings up an interesting point. It seems like it's very common to have a release that doesn't boot, or produces immediately obvious problems. I'm curious if you do any testing (LTP or otherwise) on the kernels you intend to release.

In the words of our fearless leader:

"regression testing"? What's that? If it compiles, it is good, if it boots up it is perfect.

It would be nice if that were true, but back here in reality things are rarely if ever even stable enough for testing if they merely build and boot.

If Linus really is building and booting every kernel prior to release, it would be quick and simple to add a fast subset of LTP to the mix and do a quick regression run. It's convenient, fast and could save a lot of headaches for a lot of people later on.

-Paul Larson

#11 lore raw
Quoted 5 lines · Paul Larson

If Linus really is building and booting every kernel prior to release, it would be quick and simple to add a fast subset of LTP to the mix and do a quick regression run. It's convenient, fast and could save a lot of headaches for a lot of people later on.

The problem I see with this approach is that "a lot of people" scales far better than Linus.

Saving 100 people a day of work by offloading it to Linus is quite an optimisation, but it doesn't optimize the overall development speed.

Let the crowd build the kernel, see the breakage and fix it up, until we get back to -preX and -rcY kernels.

Jörn

Signature · 2 lines
Victory in war is not repetitious.
-- Sun Tzu
#12 lore raw
In reply to Paul Larson
Quoted 8 lines · Paul Larson

It would be nice if that were true, but back here in reality things are rarely if ever even stable enough for testing if they merely build and boot.

If Linus really is building and booting every kernel prior to release, it would be quick and simple to add a fast subset of LTP to the mix and do a quick regression run. It's convenient, fast and could save a lot of headaches for a lot of people later on.

Nothing stops people from LTPtesting the -bk nightlies. Sure, they won't catch the last-minute-torvalds-breaks-the-compile type bugs, but for the most part it should be useful enough info.

Dave

Signature · 2 lines
| Dave Jones.        http://www.codemonkey.org.uk
| SuSE Labs
#13 lore raw
Quoted 3 lines · Dave Jones

Nothing stops people from LTPtesting the -bk nightlies. Sure, they won't catch the last-minute-torvalds-breaks-the-compile type bugs, but for the most part it should be useful enough info.

Already been doing that for a long time now. How about a quick note out to lkml that says "The current bk is what I'm going to release at <NN Time> today unless someone gives me a good reason not to."?

-Paul Larson

#14 lore raw
Quoted 6 lines
Quoted 3 lines

Nothing stops people from LTPtesting the -bk nightlies. Sure, they won't catch the last-minute-torvalds-breaks-the-compile type bugs, but for the most part it should be useful enough info.

Already been doing that for a long time now. How about a quick note out to lkml that says "The current bk is what I'm going to release at <NN Time> today unless someone gives me a good reason not to."?

Why? That would just delay releases, and make more work for Linus.

If a release is badly broken, another one is usually quick to follow it, anyway.

John.

#15 lore raw
Quoted 8 lines · John Bradford
Quoted 6 lines
Quoted 3 lines

Nothing stops people from LTPtesting the -bk nightlies. Sure, they won't catch the last-minute-torvalds-breaks-the-compile type bugs, but for the most part it should be useful enough info.

Already been doing that for a long time now. How about a quick note out to lkml that says "The current bk is what I'm going to release at <NN Time> today unless someone gives me a good reason not to."?

Why? That would just delay releases, and make more work for Linus.

What I just suggested would be a short 1 line note to lkml. I know he's very busy, but what's that, like 10 seconds?

Quoted 2 lines

If a release is badly broken, another one is usually quick to follow it, anyway.

There's usually a lag of 30min to an hour between the last changeset and the the one that changes the version tag anyway. I would hope/assume(dangerous) this is when it's beeing built and tested. One more script to that mix that runs a subset of ltp might add an additional 5 min. Alternatively, a note of intent to lkml might add a few seconds to that delay.

If I counted timezones etc. right, here's a quick picture of the number of minutes between the last changeset and the changeset that tagged it with the version number:
2.5.60 52 min.
2.5.59 42 min.
2.5.58 31 min.
2.5.57 16 min.
*** 2.5.58 was release something like 12 hours later

Is it less work to do a few minutes of extra testing, or go through another release in the same day?

-Paul Larson

#16 lore raw
Quoted 6 lines
Quoted 4 lines
Quoted 3 lines

How about a quick note out to lkml that says "The current bk is what I'm going to release at <NN Time> today unless someone gives me a good reason not to."?

Why? That would just delay releases, and make more work for Linus.

What I just suggested would be a short 1 line note to lkml. I know he's very busy, but what's that, like 10 seconds?

10 seconds, plus the time waiting around for replies, and the time spent reading the replies.

Quoted 18 lines
Quoted 2 lines

If a release is badly broken, another one is usually quick to follow it, anyway.

There's usually a lag of 30min to an hour between the last changeset and the the one that changes the version tag anyway. I would hope/assume(dangerous) this is when it's beeing built and tested. One more script to that mix that runs a subset of ltp might add an additional 5 min. Alternatively, a note of intent to lkml might add a few seconds to that delay.

If I counted timezones etc. right, here's a quick picture of the number of minutes between the last changeset and the changeset that tagged it with the version number:
2.5.60 52 min.
2.5.59 42 min.
2.5.58 31 min.
2.5.57 16 min.
*** 2.5.58 was release something like 12 hours later

Quoted 2 lines

Is it less work to do a few minutes of extra testing, or go through another release in the same day?

Probably less work for Linus to go through another release, plus it means that people who are not testing -bk snapshots for whatever reason are more involved in the development process.

John.

#17 lore raw
Quoted 9 lines · John Bradford
Quoted 6 lines
Quoted 4 lines
Quoted 3 lines

How about a quick note out to lkml that says "The current bk is what I'm going to release at <NN Time> today unless someone gives me a good reason not to."?

Why? That would just delay releases, and make more work for Linus.

What I just suggested would be a short 1 line note to lkml. I know he's very busy, but what's that, like 10 seconds?

10 seconds, plus the time waiting around for replies, and the time spent reading the replies.

Ideally, there should be no waiting around for replies. The message is sent, he starts whatever build/boot test cycle, checks for replies when he's done and ready to release. If nothing looks urgent enough to hold it up, then he pushes the release. I still don't see how this adds any kind of terrible delay.

-Paul Larson

#18 lore raw
Quoted 5 lines · Paul Larson

Ideally, there should be no waiting around for replies. The message is sent, he starts whatever build/boot test cycle, checks for replies when he's done and ready to release. If nothing looks urgent enough to hold it up, then he pushes the release. I still don't see how this adds any kind of terrible delay.

Outside suggestions to "improve" Linus's workflow usually fall upon deaf ears...

IMO to accomplish your goals, set up a test box with BitKeeper, constantly pulling and testing the latest 2.5.x BK trees. If they crash, send full info to lkml.

Enough crash messages, and people will know automatically whether or not the kernel is good... and Linus didn't have to be bothered at all.

Jeff

#19 lore raw
Quoted 8 lines
Quoted 5 lines

Ideally, there should be no waiting around for replies. The message is sent, he starts whatever build/boot test cycle, checks for replies when he's done and ready to release. If nothing looks urgent enough to hold it up, then he pushes the release. I still don't see how this adds any kind of terrible delay.

Outside suggestions to "improve" Linus's workflow usually fall upon deaf ears...

The only suggestions really worth making, in my opinion, are whether it's easier to work with more frequent, smaller releases, or less frequent larger releases.

Infact, it's more or less directly analogous to packet size in a file transfer protocol - fewer larger packets are more efficient, unless there is a problem, when you have to re-transmit a whole, large, packet.

Likewise with kernel releases - fewer, larger releases work fine and mean less effort for developers, unless something breaks, in which case there is a lot to go through to locate the problem, and people who can't boot the broken kernel have to wait longer to test other things that were newly merged in that release.

John.

#20 lore raw
Quoted 5 lines · John Bradford

Likewise with kernel releases - fewer, larger releases work fine and mean less effort for developers, unless something breaks, in which case there is a lot to go through to locate the problem, and people who can't boot the broken kernel have to wait longer to test other things that were newly merged in that release.

This was exactly what I was getting at. I suspect that there are a good number of people that try to boot a 2.5 kernel for testing, run into immediate problems, and shelve the idea of 2.5 testing for a couple of months because of an immediate appearance that 2.5 is too unstable to test. I've seen frequent griping that not enough testing happens, the idea is to get it to a point where more people can test it _without_ adding a huge delay or making a huge gap between releases.

Quoted 2 lines · Jeff Garzik

Outside suggestions to "improve" Linus's workflow usually fall upon deaf ears...

Since Linus hasn't chimed in yet, I'm guessing that's exactly what happened. I'm not trying to improve his workflow, but rather the workflow of anyone who might be interested in getting more involved in 2.5 testing.

-Paul Larson

#21 lore raw
Quoted 5 lines
Quoted 5 lines

Likewise with kernel releases - fewer, larger releases work fine and mean less effort for developers, unless something breaks, in which case there is a lot to go through to locate the problem, and people who can't boot the broken kernel have to wait longer to test other things that were newly merged in that release.

Quoted 7 lines

This was exactly what I was getting at. I suspect that there are a good number of people that try to boot a 2.5 kernel for testing, run into immediate problems, and shelve the idea of 2.5 testing for a couple of months because of an immediate appearance that 2.5 is too unstable to test. I've seen frequent griping that not enough testing happens, the idea is to get it to a point where more people can test it _without_ adding a huge delay or making a huge gap between releases.

I can see what you mean, but realistically, I don't see how it's practical.

You can always use 2.5.X-BK1 to get the fixes that we would probably have been in 2.5.X if Linus had done more extensive testing on it before releasing it.

John.

#22 lore raw
In reply to Paul Larson
Quoted 4 lines · Paul Larson

Since Linus hasn't chimed in yet, I'm guessing that's exactly what happened. I'm not trying to improve his workflow, but rather the workflow of anyone who might be interested in getting more involved in 2.5 testing.

What would help a lot of people (certainly me, at least), would be if somebody kept a well-publicized "already known errata" list along with (possibly unofficial) work-around patches. Something along the line of:

compile fails in drivers/widget/fooby.c with error:
undefined structure member 'blat' in line 1149.
To fix: apply <this patch>

Quoted 3 lines · John Bradford

You can always use 2.5.X-BK1 to get the fixes that we would probably have been in 2.5.X if Linus had done more extensive testing on it before releasing it.

Almost but not quite what I meant - unless -BK1 is reserved for after-release
whoops and doesn't contain "new stuff for release N+1". If -BK1 is only
bugfixes, that would be good.

Signature · 3 lines
				Valdis Kletnieks
				Computer Systems Senior Engineer
				Virginia Tech
#23 lore raw
Quoted 7 lines · On Thu, 2003-02-13 at 16:20, [email protected] wrote:

What would help a lot of people (certainly me, at least), would be if somebody kept a well-publicized "already known errata" list along with (possibly unofficial) work-around patches. Something along the line of:

compile fails in drivers/widget/fooby.c with error:
undefined structure member 'blat' in line 1149.
To fix: apply <this patch>

Good idea, I'd bet that might cut down on some of the duplicate bug reports too (at least the annoying ones that occur days or even weeks after the initial report).

I certainly wouldn't mind throwing something like this up on the LTP website if people are interested. If anyone wants to spam me with some initial data for 2.5.60 feel free.

-Paul Larson

#24 lore raw
In reply to [email protected]

On Thu, 13 Feb 2003 17:20:55 -0500
[email protected] wrote:

| On Thu, 13 Feb 2003 15:57:56 CST, Paul Larson said:
| 
| > Since Linus hasn't chimed in yet, I'm guessing that's exactly what
| > happened.  I'm not trying to improve his workflow, but rather the
| > workflow of anyone who might be interested in getting more involved in
| > 2.5 testing.
| 
| What would help a lot of people (certainly me, at least), would be if
| somebody kept a well-publicized "already known errata" list along with
| (possibly unofficial) work-around patches.  Something along the line of:
| 
| compile fails in drivers/widget/fooby.c with error:
| undefined structure member 'blat' in line 1149.
| To fix:   apply <this patch>

Yes, I agree, that would be helpful.

I try to keep current lkml/etc patches for fixes/cleanups to the latest kernel, and I've thought about a way to post them, but it's too time-consuming a task, especially when it's not one's job to do that.

| On Thu, 13 Feb 2003 22:11:17 +0000 (GMT), John Bradford said:
| 
| > You can always use 2.5.X-BK1 to get the fixes that we would probably
| > have been in 2.5.X if Linus had done more extensive testing on it
| > before releasing it.
| 
| Almost but not quite what I meant - unless -BK1 is reserved for after-release
| whoops and doesn't contain "new stuff for release N+1".   If -BK1 is only
| bugfixes, that would be good.  

--
~Randy

#25 lore raw
Quoted 6 lines · Randy.Dunlap

Yes, I agree, that would be helpful.

I try to keep current lkml/etc patches for fixes/cleanups to the latest kernel, and I've thought about a way to post them, but it's too time-consuming a task, especially when it's not one's job to do that.

We'll see how time-consuming it is, I'll try my best to keep up!

Could you send me what you have collected for 2.5.60? I've set up a placeholder on the ltp website, you can either go to ltp.sf.net and click on "Kernel Errata" or the direct URL is
http://ltp.sourceforge.net/errata but there's nothing there yet. I'll try to put some data out there tonight.

Thanks,
Paul Larson

#26 lore raw
In reply to [email protected]
Quoted 5 lines
Quoted 4 lines

Since Linus hasn't chimed in yet, I'm guessing that's exactly what happened. I'm not trying to improve his workflow, but rather the workflow of anyone who might be interested in getting more involved in 2.5 testing.

Quoted 7 lines

What would help a lot of people (certainly me, at least), would be if somebody kept a well-publicized "already known errata" list along with (possibly unofficial) work-around patches. Something along the line of:

compile fails in drivers/widget/fooby.c with error:
undefined structure member 'blat' in line 1149.
To fix: apply <this patch>

Well, you can do that with my bug database - just open a new bug report for each compile failiure, upload a patch to the database, and link the patch with the new bug. You can then collect together all the relevant bug reports in to a confirmed bug for each kernel version. You don't even have to duplicate bug reports if they occur in multiple kernel versions.

John.

#28 lore raw
In reply to Jeff Garzik

Sorry about getting back on the thread late was off doing boring management stuff.

But this is what PLM/STP does but right now it doesn't bother to send the results to any list.

http://www.osdl.org/projects/26lnxstblztn/results/

Tim

Quoted 27 lines · Jeff Garzik
Quoted 5 lines · Paul Larson

Ideally, there should be no waiting around for replies. The message is sent, he starts whatever build/boot test cycle, checks for replies when he's done and ready to release. If nothing looks urgent enough to hold it up, then he pushes the release. I still don't see how this adds any kind of terrible delay.

Outside suggestions to "improve" Linus's workflow usually fall upon deaf ears...

IMO to accomplish your goals, set up a test box with BitKeeper, constantly pulling and testing the latest 2.5.x BK trees. If they crash, send full info to lkml.

Enough crash messages, and people will know automatically whether or not the kernel is good... and Linus didn't have to be bothered at all.

Jeff

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

Signature · 5 lines
Timothy D. Witham - Lab Director - [email protected]
Open Source Development Lab Inc - A non-profit corporation
15275 SW Koll Parkway - Suite H - Beaverton OR, 97006
(503)-626-2455 x11 (office)    (503)-702-2871     (cell)
(503)-626-2436     (fax)
#29 lore raw
Quoted 7 lines · Timothy D. Witham

Sorry about getting back on the thread late was off doing boring management stuff.

But this is what PLM/STP does but right now it doesn't bother to send the results to any list.

http://www.osdl.org/projects/26lnxstblztn/results/

Neat, thanks for posting the link.

IMO, it would be nice to send results to linux-kernel, but with a few restrictions:

* just a small email, with only key bits of info.
URLs would point to more detailed information.
* a constant URL, which describes what the heck the email is all about (such as your above URL)
* for now, only bother with "ia32 Default"
* never email more than once a day... even if the bot gets stuck in a spamming loop, you need to have something in place to throttle emails.
* only email when state changes: i.e. PASS->FAIL or FAIL->PASS, never PASS->PASS or FAIL->FAIL. [debateable... some may disagree with me on this one]

Comments?

Jeff

P.S. The column headers in your report are broken, for example "ia32 Default" goes to bad link
http://www.osdl.org/projects/plm/def_ia32_default.html

#30 lore raw
Quoted 14 lines · Jeff Garzik
Quoted 7 lines · Timothy D. Witham

Sorry about getting back on the thread late was off doing boring management stuff.

But this is what PLM/STP does but right now it doesn't bother to send the results to any list.

http://www.osdl.org/projects/26lnxstblztn/results/

Neat, thanks for posting the link.

IMO, it would be nice to send results to linux-kernel, but with a few restrictions:

Quoted 11 lines

* just a small email, with only key bits of info.
URLs would point to more detailed information.
* a constant URL, which describes what the heck the email is all about (such as your above URL)
* for now, only bother with "ia32 Default"
* never email more than once a day... even if the bot gets stuck in a spamming loop, you need to have something in place to throttle emails.
* only email when state changes: i.e. PASS->FAIL or FAIL->PASS, never PASS->PASS or FAIL->FAIL. [debateable... some may disagree with me on this one]

I'm not a big fan of only on transition as sometimes things
get broken but nothing ever gets reported. But your suggestions will be taken and implemented.

Quoted 8 lines

Comments?

Jeff

P.S. The column headers in your report are broken, for example "ia32 Default" goes to bad link
http://www.osdl.org/projects/plm/def_ia32_default.html

Signature · 5 lines
Timothy D. Witham - Lab Director - [email protected]
Open Source Development Lab Inc - A non-profit corporation
15275 SW Koll Parkway - Suite H - Beaverton OR, 97006
(503)-626-2455 x11 (office)    (503)-702-2871     (cell)
(503)-626-2436     (fax)
#31 lore raw
In reply to Paul Larson
Quoted 22 lines
Quoted 1 line

Why? That would just delay releases, and make more work for Linus.

What I just suggested would be a short 1 line note to lkml. I know he's very busy, but what's that, like 10 seconds?

Quoted 2 lines

If a release is badly broken, another one is usually quick to follow it, anyway.

There's usually a lag of 30min to an hour between the last changeset and the the one that changes the version tag anyway. I would hope/assume(dangerous) this is when it's beeing built and tested. One more script to that mix that runs a subset of ltp might add an additional 5 min. Alternatively, a note of intent to lkml might add a few seconds to that delay.

If I counted timezones etc. right, here's a quick picture of the number of minutes between the last changeset and the changeset that tagged it with the version number:
2.5.60 52 min.
2.5.59 42 min.
2.5.58 31 min.
2.5.57 16 min.
*** 2.5.58 was release something like 12 hours later

Quoted 2 lines

Is it less work to do a few minutes of extra testing, or go through another release in the same day?

Or you just persaude Linus to release kernels first thing in the morning, not last thing at night. Then the automatic nightly tests would cover it ...(assuming you wake up before he does, and email him the results ... but you're 2hrs earlier than him, so ... ;-))

M.

#32 lore raw
Quoted 3 lines · Martin J. Bligh

Or you just persaude Linus to release kernels first thing in the morning, not last thing at night.

Ok, now the thread has moved from the strange to the surreal.

"First thing in the morning" indeed.

Trust me, you don't want me doing _anything_ first thing in the morning.

Linus

#33 lore raw
Quoted 3 lines · Linus Torvalds

Trust me, you don't want me doing _anything_ first thing in the morning.

Linus

Doesn't Alan Cox go to bed at first thing in the morning?

Mike

#34 lore raw
In reply to Linus Torvalds
Quoted 4 lines
Quoted 2 lines

Or you just persaude Linus to release kernels first thing in the morning, not last thing at night.

Ok, now the thread has moved from the strange to the surreal.

Fed up with all these people trying to tell you when to release new kernels? Now you can just use this simple script to do it automatically!

#!/bin/sh
echo "Release a new kernel now!!!" > /tmp/kernel_reminder
echo "30 6 * * * /usr/bin/cat /tmp/kernel_reminder | mail root" >> /var/spool/cron/crontabs/root

John.

#35 lore raw
In reply to John Bradford
Quoted 2 lines · John Bradford

If a release is badly broken, another one is usually quick to follow it, anyway.

It is my impression that this is frequently not the case.

I'd love Linus to release x.y.z before he goes to bed, go through the bug reports that have trickled in while he's in a foul mood in the morning, and then make a x.y.z+1 release with the most glaring bugs fixed.

Of course, there are probably good reasons why Linus doesn't do this already.

- Werner

Signature · 3 lines
  _________________________________________________________________________
 / Werner Almesberger, Buenos Aires, Argentina         [email protected] /
/_http://www.almesberger.net/____________________________________________/
#36 lore raw
In reply to Dave Jones
Quoted 5 lines · Dave Jones
Quoted 4 lines · Paul Larson

It would be nice if that were true, but back here in reality things are rarely if ever even stable enough for testing if they merely build and boot.

Quoted 8 lines
Quoted 4 lines

If Linus really is building and booting every kernel prior to release, it would be quick and simple to add a fast subset of LTP to the mix and do a quick regression run. It's convenient, fast and could save a lot of headaches for a lot of people later on.

Nothing stops people from LTPtesting the -bk nightlies. Sure, they won't catch the last-minute-torvalds-breaks-the-compile type bugs, but for the most part it should be useful enough info.

Agreed... at least the past few releases, the just-after-the-release BK snapshot often compiles and boots more reliably than the release <g>

Current 2.5.60-BK is looking _really_ nice, LTP-wise.

Jeff