Skip to contentSkip to conversation

Recent Siano patches - testing required

DISCUSSION Uri Shkolnik 5 participants 7 messages May 19, 2009 – May 20, 2009 Media (V4L/DVB)

lore
Share & tools
lore
#0 lore raw
Quoted 29 lines · Steven Toth

From: Steven Toth <[email protected]>
Subject: Recent Siano patches - testing required
To: [email protected], "Mauro Carvalho Chehab" <[email protected]>
Cc: "[email protected]" <[email protected]>
Date: Tuesday, May 19, 2009, 10:15 PM
Mauro please review.

Uri,

Firstly I'd like to thank you and Siano for patching and helping to maintain the driver. :)

Second, this is a heck of a lot of change for the list to review! It's impossible to digest the level of rework and potential regressions.

I'd suggest you either host your own mercurial server and have testers pull your trees, helping to regression test your changes or ... someone gives you access to create trees at LinuxTV.org, then you can solicit testers feedback on the mailing list.

Either way, it's unusual for this amount of change to be merged without having some positive feedback from the Linux community of testers. If you have confirmation that all of the current devices are still working correct, without regression, then please indicate this in your patches / email.

Quoted 6 lines

If not, the patches should be hosted somewhere for test and review.

-- Steven Toth - Kernel Labs
http://www.kernellabs.com

Steven,

Thanks for your comments.

Brief history about Siano's Linux kernel drivers...

First set of Siano's based receivers drivers has been introduced at the first half of 2008.

Since mid-2008 till about little more than a month ago, from various reasons, (most of these reasons were unrelated to Siano's intentions), there were no submissions to the LinuxTV mercurial.

However, lots of progress with Linux drivers has been achieved. Some of it within the Linux various communities (for example with the MMC maintainer, and members of this mailing list based on direct contacts) but primarily with Siano's commercial customers which based their products on Linux OS or it derivatives (such as Google's Android).

Note that the recent month' patches came to bring the LinuxTV mercurial (and kernel's git as a result) up to the state of Siano's internal Subversion repository. Meaning, the patches are not new code / new bugs fixes etc. but rather up-stream from Siano's repository to LinuxTV's (same as it done from each sub-kernel system (including LinuxTV) to the kernel git).

The Siano's drivers have been tested thoroughly, both in Siano's QA departments (where a full time QA engineer is tasked to test various Linux/SMS based setups) and in customers sites as well.

Note that the current Siano's Linux drivers set is been used by many dozens of commercial products (including tear-1 companies' products), which of course have been tested thoroughly that offering, and it is a production level code (literally).

The risks as I see them with all those patches, are that some work (minor) has been done in order to make these patches appropriate to be submitted (per all rules of patches submission to LinuxTV / Linux kernel).

Another problem is that I find it hard to believe that there will be many testers available from this mailing list. True that there are some dongle and mini-cards based on Siano's chip-set which these devices' manufacturers representatives are on this list, orsome people may own these, but the vast majority of Siano's based products are embedded based devices (including multiple tablet/UMPC PCs, cellular phones, PDAs, navigation devices, DVD/BR players, gaming devices and many others), so unless someone will hack those devices and will replace the installed drivers (kernel image actually, since most of these fixed-targets developers build the Siano's modules to be included within the kernel image) there is no much hope for comprehensive o"open testers" test.

However any test that will be performed, will benefit all (including Siano.... :-)

It's true that is the recent past, Siano equipped some volunteers from this community with devices (free of charge), but it had been done base on Siano's "selfish" objectives, which at the time suited various individuals. I find it hard to believe that Siano will equip anyone just for "random" testing, since we have enough brimful testing environments.

Best Regards,

Uri

#1 lore raw
Quoted 2 lines

However any test that will be performed, will benefit all (including Siano.... :-)

Agreed. Yes, I happen to know Hauppauge very well.

I'm very happy to see that the driver is being improved but I'll be even happier to see actual testers report success before any of this stuff is merged. My concern is the vast amount of change coming through this list and expected to be merged blindly into the kernel.

If we have no testers then, at least for Hauppauge products, we'll find some. Let me know if I can help with this.

Until then nothing should be blindly merged that could regress existing product support.

Mauro?

Signature · 2 lines
Steven Toth - Kernel Labs
http://www.kernellabs.com
#2 lore raw

Em Tue, 19 May 2009 17:30:42 -0400
Steven Toth <[email protected]> escreveu:

Quoted 10 lines
Quoted 2 lines

However any test that will be performed, will benefit all (including Siano.... :-)

Agreed. Yes, I happen to know Hauppauge very well.

I'm very happy to see that the driver is being improved but I'll be even happier to see actual testers report success before any of this stuff is merged. My concern is the vast amount of change coming through this list and expected to be merged blindly into the kernel.

Quoted 7 lines

If we have no testers then, at least for Hauppauge products, we'll find some. Let me know if I can help with this.

Until then nothing should be blindly merged that could regress existing product support.

Mauro?

Steven,

Your concerns about testing make sense, but this were already tried in the past, when Uri started sending their patches at the ML. So, instead of repeating the same novel, let's merge the patches at the development tree and ask people to test.

Yet, I'm keeping the Siano patches at the 'pending' -git tree, where I hold very experimental work. I intend to hold it there until we have more tests and have all the pending patches merged.

About creating an -hg tree for Siano (and for other developers), I had to nack it in the past, since the LinuxTV server were overloaded. Now that the machine got replaced, I think we may actually create a tree for them.

Uri, please discuss about this in priv with me, in order to exchange the needed information for the login account.

Cheers,
Mauro

#3 lore raw
Quoted 15 lines · Mauro Carvalho Chehab

Steven,

Your concerns about testing make sense, but this were already tried in the past, when Uri started sending their patches at the ML. So, instead of repeating the same novel, let's merge the patches at the development tree and ask people to test.

Yet, I'm keeping the Siano patches at the 'pending' -git tree, where I hold very experimental work. I intend to hold it there until we have more tests and have all the pending patches merged.

About creating an -hg tree for Siano (and for other developers), I had to nack it in the past, since the LinuxTV server were overloaded. Now that the machine got replaced, I think we may actually create a tree for them.

Quoted 7 lines

Uri, please discuss about this in priv with me, in order to exchange the needed information for the login account.

Cheers,
Mauro

Mauro,

If I recall, a ton of patches were sent to the mailing list, but there was never a test tree on linuxtv.org. Were you expecting interested parties to hand-apply all those patches?

It's not clear to me why you are putting this code that is untested by the community into the v4l-dvb tree. In all other cases where linuxtv developers want to submit large sets of changes, you expect them to create a private tree so testers can be solicited *before* it goes into v4l-dvb. Why would this case be any different?

You have no personal knowledge as to whether the code actually works, and there are parties who have expressed a very clear concern about some of the patches causing regressions in existing hardware. Doesn't the safer approach seem to be to setup a ~mchehab/siano-patches tree so that people can do some testing, and *then* merge to the mainline? If everything works, then great - it's one command to merge the fold the tree back into the mainline...

Devin

Signature · 2 lines
Devin J. Heitmueller - Kernel Labs
http://www.kernellabs.com
#4 lore raw

Hi,

Am Dienstag, den 19.05.2009, 22:05 -0400 schrieb Devin Heitmueller:

Quoted 45 lines
Quoted 15 lines · Mauro Carvalho Chehab

Steven,

Your concerns about testing make sense, but this were already tried in the past, when Uri started sending their patches at the ML. So, instead of repeating the same novel, let's merge the patches at the development tree and ask people to test.

Yet, I'm keeping the Siano patches at the 'pending' -git tree, where I hold very experimental work. I intend to hold it there until we have more tests and have all the pending patches merged.

About creating an -hg tree for Siano (and for other developers), I had to nack it in the past, since the LinuxTV server were overloaded. Now that the machine got replaced, I think we may actually create a tree for them.

Quoted 7 lines

Uri, please discuss about this in priv with me, in order to exchange the needed information for the login account.

Cheers,
Mauro

Mauro,

If I recall, a ton of patches were sent to the mailing list, but there was never a test tree on linuxtv.org. Were you expecting interested parties to hand-apply all those patches?

It's not clear to me why you are putting this code that is untested by the community into the v4l-dvb tree. In all other cases where linuxtv developers want to submit large sets of changes, you expect them to create a private tree so testers can be solicited *before* it goes into v4l-dvb. Why would this case be any different?

You have no personal knowledge as to whether the code actually works, and there are parties who have expressed a very clear concern about some of the patches causing regressions in existing hardware. Doesn't the safer approach seem to be to setup a ~mchehab/siano-patches tree so that people can do some testing, and *then* merge to the mainline? If everything works, then great - it's one command to merge the fold the tree back into the mainline...

Quoted 2 lines

Devin

unfortunately, IIRC, Linus said once testing is for idiots or something like that only.

He can say that :)

But here the rule is, to allow him further to say so, you are not allowed to send in even a oneliner only without compile and functional test :)

As Mauro explained, Uri will likely very soon have his public test repo.

Cheers,
Hermann

#5 lore raw
In reply to Devin Heitmueller

Em Tue, 19 May 2009 22:05:19 -0400
Devin Heitmueller <[email protected]> escreveu:

Quoted 5 lines

It's not clear to me why you are putting this code that is untested by the community into the v4l-dvb tree. In all other cases where linuxtv developers want to submit large sets of changes, you expect them to create a private tree so testers can be solicited *before* it goes into v4l-dvb. Why would this case be any different?

Because it was the way we've indicated to Uri for submitting his patches:

http://www.mail-archive.com/[email protected]/msg00473.html

We may change it for the next patch series, but I'll pick the good patches from this series via patchwork.

Cheers,
Mauro

#6 lore raw
In reply to Mauro Carvalho Chehab
Quoted 2 lines

Uri, please discuss about this in priv with me, in order to exchange the needed information for the login account.

Thank you Mauro.

Signature · 2 lines
Steven Toth - Kernel Labs
http://www.kernellabs.com