Bug#884154: [Pkg-javascript-devel] Bug#884154: FTBFS with chai 4.1.2 in experimental

[Pirate Praveen]

On Sat, 16 Dec 2017 17:14:07 + Ghislain Vaillant
 wrote> I don't know anything about chai. Besides,
since you are its maintainer,
> I would expect the investigation to be done by yourself rather than myself.

With help from Aruna, I was able to fix all other packages and now this
is the only package blocking a chai 4 upload to unstable. She was not
able to reproduce this bug, also because of #887583 I can't confirm the
failure again.

See #887583 you did not even get a heads up when mocha was updated :(



signature.asc
Description: OpenPGP digital signature

Bug#884154: [Pkg-javascript-devel] Bug#884154: FTBFS with chai 4.1.2 in experimental

[Pirate Praveen]

On ശനി 16 ഡിസംബര്‍ 2017 11:16 വൈകു, Ghislain Vaillant wrote:
> You are deliberately taking an extreme case to fulfill your narrative. I
> would not fit nodejs and chai in the same basket.

see below, ruby-webmock is comparable here.

>> I did not even get heads up with webmock transition in ruby team. It was
>> right away FTBFS and autoremoval from testing for packages failing to
>> build with webmock 3.
> 
> And?

I had to get the tests ported to ruby-webmock 3, not the person who
uploaded ruby-webmock 3.

3 of my packages got removed from testing.

>> I can try and help, but as maintainer of libjs-fetch, it is your
>> responsibility to fix issues of your package when dependencies change.
> 
> We are talking about a *test* dependencies which now makes the build
> fail after a major version bump. It makes sense to expect more
> information from the corresponding maintainer.

That is how ruby-webmock, ruby-rspec etc were handled.

> Should we expect every single maintainer affected by an FTBFS to go read
> the release notes of chai in order to figure what broke between version
> 3 and 4. I am very skeptical about this.

yes, at least that is how I had to do with many ruby test framework
transitions and I think that is only reasonable or we can't really
update any package that is depended on by large number of packages.

>> 1. chai itself is FTBFS with nodejs 6
> 
> That's unfortunate indeed.
> 
>> 2. we generally want to ship the latest versions
> 
> But not systematically.
> 
>> 3. Some packages are already starting to require newer versions of chai,
>> for example node-yargs (whose tests are disabled currently).
> 
> It makes sense to disable them if they specifically require chai >= 4.
> 

Right now I did that, but going forward, it makes to sense to
disable/port tests of packages that needs an older version. That is how
any transition is handled, chai is not anything special. I know it is
hard and I have to do the same for my other packages too, but that is
part of what it takes to maintain a package.

>> You could ask upstream to move to chai 4 and take their help. Or you
>> could also disable tests.
> 
> Disabling the tests would be a serious downgrade considering testing is
> currently working, unlike node-yargs. I am seriously uncomfortable with
> this proposal.

node-yargs tests are working with chai 4 and we are in a better to
position to support chai 4 than chai 3.5 for buster.

> I can ask upstream, though chai is officially pinned at version 2.x there.
> 

I do that regularly. I did that for ruby-webmock 3.



signature.asc
Description: OpenPGP digital signature

Bug#884154: [Pkg-javascript-devel] Bug#884154: FTBFS with chai 4.1.2 in experimental

[Ghislain Vaillant]

Le 16/12/17 à 17:31, Pirate Praveen a écrit :

On ശനി 16 ഡിസംബര്‍ 2017 10:44 വൈകു, Ghislain Vaillant wrote:

I don't know anything about chai. Besides, since you are its maintainer,
I would expect the investigation to be done by yourself rather than myself.


No, that is not how a transition is supposed to work. Do you expect
maintainer of nodejs package to fix all issues of all packages using
nodejs by themselves when updating nodejs versions?


You are deliberately taking an extreme case to fulfill your narrative. I 
would not fit nodejs and chai in the same basket.



I did not even get heads up with webmock transition in ruby team. It was
right away FTBFS and autoremoval from testing for packages failing to
build with webmock 3.


And?


I can try and help, but as maintainer of libjs-fetch, it is your
responsibility to fix issues of your package when dependencies change.


We are talking about a *test* dependencies which now makes the build 
fail after a major version bump. It makes sense to expect more 
information from the corresponding maintainer.


Should we expect every single maintainer affected by an FTBFS to go read 
the release notes of chai in order to figure what broke between version 
3 and 4. I am very skeptical about this.



I meant context about why the package now FTBFS. I understand this is a
transition and I don't think uploading to unstable is wise before FTBFS
issues such as this one are fixed. Is there an urgency in having chai
4.x in testing/unstable?


Indeed, that is why it is uploaded first to experimental and giving
heads up to packages affected by this transition. We can definitely wait
for a reasonable time before uploading to unstable.


Which is appreciated.


1. chai itself is FTBFS with nodejs 6


That's unfortunate indeed.


2. we generally want to ship the latest versions


But not systematically.


3. Some packages are already starting to require newer versions of chai,
for example node-yargs (whose tests are disabled currently).


It makes sense to disable them if they specifically require chai >= 4.


You could ask upstream to move to chai 4 and take their help. Or you
could also disable tests.


Disabling the tests would be a serious downgrade considering testing is 
currently working, unlike node-yargs. I am seriously uncomfortable with 
this proposal.


I can ask upstream, though chai is officially pinned at version 2.x there.

Bug#884154: [Pkg-javascript-devel] Bug#884154: FTBFS with chai 4.1.2 in experimental

[Pirate Praveen]

On ശനി 16 ഡിസംബര്‍ 2017 10:44 വൈകു, Ghislain Vaillant wrote:
> I don't know anything about chai. Besides, since you are its maintainer,
> I would expect the investigation to be done by yourself rather than myself.

No, that is not how a transition is supposed to work. Do you expect
maintainer of nodejs package to fix all issues of all packages using
nodejs by themselves when updating nodejs versions?

I did not even get heads up with webmock transition in ruby team. It was
right away FTBFS and autoremoval from testing for packages failing to
build with webmock 3.

I can try and help, but as maintainer of libjs-fetch, it is your
responsibility to fix issues of your package when dependencies change.

> I meant context about why the package now FTBFS. I understand this is a
> transition and I don't think uploading to unstable is wise before FTBFS
> issues such as this one are fixed. Is there an urgency in having chai
> 4.x in testing/unstable?

Indeed, that is why it is uploaded first to experimental and giving
heads up to packages affected by this transition. We can definitely wait
for a reasonable time before uploading to unstable.

1. chai itself is FTBFS with nodejs 6
2. we generally want to ship the latest versions
3. Some packages are already starting to require newer versions of chai,
for example node-yargs (whose tests are disabled currently).

You could ask upstream to move to chai 4 and take their help. Or you
could also disable tests.



signature.asc
Description: OpenPGP digital signature

Bug#884154: [Pkg-javascript-devel] Bug#884154: FTBFS with chai 4.1.2 in experimental

[Ghislain Vaillant]

Le 16/12/17 à 17:07, Pirate Praveen a écrit :

On ശനി 16 ഡിസംബര്‍ 2017 08:41 വൈകു, Ghislain Vaillant wrote:

Why is this happening with chai 4.x and not 3.5?


That is what we need to find out. Possibly some deprecated methods are
removed from new version.


I don't know anything about chai. Besides, since you are its maintainer, 
I would expect the investigation to be done by yourself rather than myself.



Sorry, but I am gonna need more context to fix this.


We are trying to update chai from 3.5 to 4.x but noticed this package
fails with chai 4.x. Once we upload chai 4.x to unstable, it will cause
FTBFS (and severity will be changed to serious) and we wanted to give a
heads up before that. This is the normal procedure of a transition.


I meant context about why the package now FTBFS. I understand this is a 
transition and I don't think uploading to unstable is wise before FTBFS 
issues such as this one are fixed. Is there an urgency in having chai 
4.x in testing/unstable?

Bug#884154: [Pkg-javascript-devel] Bug#884154: FTBFS with chai 4.1.2 in experimental

[Pirate Praveen]

On ശനി 16 ഡിസംബര്‍ 2017 08:41 വൈകു, Ghislain Vaillant wrote:
> Why is this happening with chai 4.x and not 3.5?

That is what we need to find out. Possibly some deprecated methods are
removed from new version.

> Sorry, but I am gonna need more context to fix this.

We are trying to update chai from 3.5 to 4.x but noticed this package
fails with chai 4.x. Once we upload chai 4.x to unstable, it will cause
FTBFS (and severity will be changed to serious) and we wanted to give a
heads up before that. This is the normal procedure of a transition.



signature.asc
Description: OpenPGP digital signature