Re: [Reproducible-builds] Usefulness of periodic reproducible builds e-mails

2015-10-05 Thread Holger Levsen
Hi Emmanuel,

On Montag, 5. Oktober 2015, Emmanuel Bourg wrote:
> Thank you very much for the changes. Let's see how it goes, but I
> suspect the volume of notifications will be much better and the need for
> individual subscriptions less critical.

we'll see. you are still subscribed to 1000 different packages, so there is 
still the potential for 1000 mails a day.
 
> Is it possible to increase the threshold to 2 notifications per package
> per day? Sometimes we fix build issues the same day they are reported,
> and getting the fix notification would be useful.

wouldnt that defeat the purpose of the limitations? Sometimes we break+fix our 
infrastructure on the same day, meaning you'd then might get twice the 
notifications.


cheers,
Holger


signature.asc
Description: This is a digitally signed message part.
___
Reproducible-builds mailing list
Reproducible-builds@lists.alioth.debian.org
http://lists.alioth.debian.org/cgi-bin/mailman/listinfo/reproducible-builds

Re: [Reproducible-builds] Usefulness of periodic reproducible builds e-mails

2015-10-02 Thread Holger Levsen
Hi Emmanuel,

On Dienstag, 29. September 2015, Emmanuel Bourg wrote:
> > I agree and am wondering if we should actually do this, and limit
> > (maintainer) notifications to unstable? What do you think?
> Well, if I understood your "this graph is a lie" properly in your talks,
> I think the reproducibility in testing isn't very interesting for now,
> until the tool chain matures and we have really reproducible packages in
> unstable. At this point the testing notifications will have much more
> sense.

Nope, I don't think you understood my "this graph is a lie" properly ;-) The 
graphs are "lies", because they don't show sid and stretch but (sid+our repo) 
and (stretch+our repo).

There is a different reason why I think notifications for testing are "useless 
noise" (or "not so interesting information", if you prefer): in Debian, we fix 
things in sid and these fixes migrate to testing (=stretch), so once a package 
has become reproducible in sid it should also become reproducible in testing, 
once that version migrates to testing. 

If this doesnt happen it's almost certainly a bug in our test framework, but 
not a reproducibility issue in the package. And if the package ftbfs in 
testing, this is very sad, but IMO not appropriate to send a "reproducible 
builds project" notification for it - such problems should be detected 
elsewhere. It's nice if we gather that data, and we should also manually file 
bugs from that data, but I dont think we should generate automated 
notifications because as I tried to explain, if a package is fixed in sid, the 
fix will migrate to testing eventually. Thus we only really need to care about 
sid and testing will be good "automatically".

> So yes, limiting maintainer notifications to unstable would be a good idea.

I've limited notifications to unstable and experimental now, and also improved 
the code a bit that only one mail per is sent per source package in all 
suites, no matter how many status changes it had. But we should still improve 
it to allow individual subscriptions, and probably this is best done via 
tracker.d.o - does anybody know how to achieve that?


cheers,
Holger


signature.asc
Description: This is a digitally signed message part.
___
Reproducible-builds mailing list
Reproducible-builds@lists.alioth.debian.org
http://lists.alioth.debian.org/cgi-bin/mailman/listinfo/reproducible-builds

Re: [Reproducible-builds] Usefulness of periodic reproducible builds e-mails

2015-09-30 Thread Markus Koschany
Hi,

Am 30.09.2015 um 12:30 schrieb Holger Levsen:
> Hi,
> 
> (mostly ignoring the rest as this has been addressed already.)
> 
> On Dienstag, 29. September 2015, Markus Koschany wrote:
>> I understand that everything is still in development. However I don't
>> think a public mailing list is a suitable testbed. My preferred solution
>> would be to make receiving such e-mails opt-in. Everyone who wants to
>> get informed by e-mail may subscribe to this feature. I personally
>> prefer and regularly check DDPO because it is quite, very informative
>> and can be used on demand. The only other option I can think of is to
>> filter pkg-java mails.
> 
> I think you still misunderstand: you, the java maintainers, already opted-in.
> 
> What I then proposed was to change the notification system to allow 
> individual 
> email addresses to be subscribed to anything (and not just packages per se as 
> it is now) and to limit the notifications to unstable.
> 
> But it still remains, that you (the team) opted in activly.
> 
> If you as a team prefer, we can easily unsubscribe you now.

We have never discussed this before as a team. I vote for unsubscribing
pkg-java because of the issues that were pointed out already.

Regards,

Markus






signature.asc
Description: OpenPGP digital signature
___
Reproducible-builds mailing list
Reproducible-builds@lists.alioth.debian.org
http://lists.alioth.debian.org/cgi-bin/mailman/listinfo/reproducible-builds

Re: [Reproducible-builds] Usefulness of periodic reproducible builds e-mails

2015-09-30 Thread Holger Levsen
Hi,

On Dienstag, 29. September 2015, Santiago Vila wrote:
> Are we really spamming people with this?
> (spam = unsolicited bulk email)

no, we dont.

please read https://reproducible.debian.net/index_notify.html


cheers,
Holger


signature.asc
Description: This is a digitally signed message part.
___
Reproducible-builds mailing list
Reproducible-builds@lists.alioth.debian.org
http://lists.alioth.debian.org/cgi-bin/mailman/listinfo/reproducible-builds

Re: [Reproducible-builds] Usefulness of periodic reproducible builds e-mails

2015-09-30 Thread Emmanuel Bourg
Le 30/09/2015 13:00, Markus Koschany a écrit :

> We have never discussed this before as a team. I vote for unsubscribing
> pkg-java because of the issues that were pointed out already.

I did request the notifications for the team, but I didn't really expect
so many false positives. Sorry for the trouble.

Markus (and the others), do you think it's ok to keep the notifications
if they are limited to unstable ? Or would you prefer disabling them
completely until the build environment stabilizes?

Emmanuel Bourg


___
Reproducible-builds mailing list
Reproducible-builds@lists.alioth.debian.org
http://lists.alioth.debian.org/cgi-bin/mailman/listinfo/reproducible-builds

Re: [Reproducible-builds] Usefulness of periodic reproducible builds e-mails

2015-09-30 Thread Miguel Landaeta
On Wed, Sep 30, 2015 at 02:27:18PM +0200, Emmanuel Bourg wrote:
> 
> Markus (and the others), do you think it's ok to keep the notifications
> if they are limited to unstable ? Or would you prefer disabling them
> completely until the build environment stabilizes?

I think there is value in receiving these notifications, they provide
timely feedback about the status of our packages.

Maybe we can setup a mailing list for this, something like
pkg-java-bui...@lists.alioth.debian.org? (we already have
pkg-java-commits, for example).

I also don't mind filtering the reproducible reports in my email
client if there is not other alternatives, of course.

-- 
Miguel Landaeta, nomadium at debian.org
secure email with PGP 0x6E608B637D8967E9 available at http://miguel.cc/key.
"Faith means not wanting to know what is true." -- Nietzsche


signature.asc
Description: Digital signature
___
Reproducible-builds mailing list
Reproducible-builds@lists.alioth.debian.org
http://lists.alioth.debian.org/cgi-bin/mailman/listinfo/reproducible-builds

Re: [Reproducible-builds] Usefulness of periodic reproducible builds e-mails

2015-09-30 Thread Markus Koschany
Am 30.09.2015 um 16:15 schrieb Mattia Rizzolo:
[...]
> This wouldn't work with the current implementation, which is emailing
> $p...@packages.debian.org.  Anyway, I received a suggestion of setting up
> a new PTS keyword, so then people can go and subscribe there, maybe
> using the team facility of the new tracker to subscribe to the whole lot
> (i beliebe it works that way?).

This is what I had in mind too. I would prefer such an implementation
over the current state. Just let interested people (humans) subscribe to
this feature with their private e-mail address.

I still think that DDPO is sufficient but if the others feel we need the
same information via e-mail then I would prefer only FTBFS reports in
unstable.

Markus



signature.asc
Description: OpenPGP digital signature
___
Reproducible-builds mailing list
Reproducible-builds@lists.alioth.debian.org
http://lists.alioth.debian.org/cgi-bin/mailman/listinfo/reproducible-builds

Re: [Reproducible-builds] Usefulness of periodic reproducible builds e-mails

2015-09-29 Thread Emmanuel Bourg
Le 29/09/2015 16:32, Markus Koschany a écrit :

> I would like to take the opportunity to raise the following concern. I
> support the reproducible builds effort but I think the periodic e-mails
> to pkg-java are often not useful enough at the moment. There are far too
> many false-positives. The list should be reserved for discussing bug
> reports and due the flood of reproducible e-mails it happens that one
> can miss a bug report.

I agree the false positive are annoying, but this is understandable
since the system is still in development (the recent addition of
disorderfs caused some troubles for example). Fortunately the
reproducible builds team is very reactive and the issues are solved
quickly. I expect the notifications to stabilize progressively.


> The reports would be most useful if they included only confirmed FTBFS
> in testing. I know detecting FTBFS is only a by-product of the
> reproducible build effort but those bugs are the most interesting ones
> for us.

Actually I tend to prefer the reports in unstable. The reports in
testing are often duplicates of issues already known in unstable, and
sometimes they are already fixed in unstable when they are notified in
testing. If we were to keep the reports for only suite I'd prefer the
unstable reports since it gives a faster feedback on our work.

Emmanuel Bourg


___
Reproducible-builds mailing list
Reproducible-builds@lists.alioth.debian.org
http://lists.alioth.debian.org/cgi-bin/mailman/listinfo/reproducible-builds

Re: [Reproducible-builds] Usefulness of periodic reproducible builds e-mails

2015-09-29 Thread Holger Levsen
Hi,

thanks for explaining, Emmanuel!

On Dienstag, 29. September 2015, Emmanuel Bourg wrote:
> Actually I tend to prefer the reports in unstable. The reports in
> testing are often duplicates of issues already known in unstable, and
> sometimes they are already fixed in unstable when they are notified in
> testing. If we were to keep the reports for only suite I'd prefer the
> unstable reports since it gives a faster feedback on our work.

I agree and am wondering if we should actually do this, and limit (maintainer) 
notifications to unstable? What do you think?


cheers,
Holger


signature.asc
Description: This is a digitally signed message part.
___
Reproducible-builds mailing list
Reproducible-builds@lists.alioth.debian.org
http://lists.alioth.debian.org/cgi-bin/mailman/listinfo/reproducible-builds

Re: [Reproducible-builds] Usefulness of periodic reproducible builds e-mails

2015-09-29 Thread Holger Levsen
Hi,

On Dienstag, 29. September 2015, Markus Koschany wrote:
> I would like to take the opportunity to raise the following concern. I
> support the reproducible builds effort but I think the periodic e-mails
> to pkg-java are often not useful enough at the moment. There are far too
> many false-positives. The list should be reserved for discussing bug
> reports and due the flood of reproducible e-mails it happens that one
> can miss a bug report.
> 
> The reports would be most useful if they included only confirmed FTBFS
> in testing. I know detecting FTBFS is only a by-product of the
> reproducible build effort but those bugs are the most interesting ones
> for us.

there are two conflicting demands here: I think the java situation is because 
there are many packages and thus if our system breaks, you get lots of false 
mails, so understandable you dont want to get false ftbfs reports. And then 
there are others, wo want these ftbfs reports, because also often they do 
indicate real problems, and if not, these other people with less packages, can 
easily mentally filter those. and then maybe (?) even you (Markus) and 
Emmanuel are on the same java team and have different preferences.

There are many possible solutions, eg we could stop mailing ftbfs reports. Or 
Markus filters them in his mail client. Or we develop a new subscription 
system, where individual users (=email addresses) can (be) subscribed, instead 
of (*) the subscribing all packages maintainers, as we do now, and which has 
caused nuicances in the past already too…


cheers,
Holger

(*) or in addition


signature.asc
Description: This is a digitally signed message part.
___
Reproducible-builds mailing list
Reproducible-builds@lists.alioth.debian.org
http://lists.alioth.debian.org/cgi-bin/mailman/listinfo/reproducible-builds

Re: [Reproducible-builds] Usefulness of periodic reproducible builds e-mails

2015-09-29 Thread Emmanuel Bourg
Le 29/09/2015 17:08, Holger Levsen a écrit :

> I agree and am wondering if we should actually do this, and limit 
> (maintainer) 
> notifications to unstable? What do you think?

Well, if I understood your "this graph is a lie" properly in your talks,
I think the reproducibility in testing isn't very interesting for now,
until the tool chain matures and we have really reproducible packages in
unstable. At this point the testing notifications will have much more sense.

So yes, limiting maintainer notifications to unstable would be a good idea.

Emmanuel Bourg


___
Reproducible-builds mailing list
Reproducible-builds@lists.alioth.debian.org
http://lists.alioth.debian.org/cgi-bin/mailman/listinfo/reproducible-builds