You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@cordova.apache.org by Joe Bowser <bo...@gmail.com> on 2013/09/18 20:35:23 UTC
Cordova JS, CordovaWebView and Coho
After I let Andrew do the tagging of RC1, I noticed something that
looks broken by the fact that I can't reproduce this result without
using coho, and I can't find in the source where coho messes with the
build labels.
Now, as well all know, the JS is generated by Grunt. Assuming that
we're going to be building off the same branch for the JS, we should
all be getting the same JS by doing this:
git checkout 3.1.0-rc1
grunt
That produces a JS file with this header:
3.1.0-rc1-0-g0d70465
However, when you look at the JS checked into Android, it's simply just this:
3.1.0-rc1
Now, they're the same, but when we remove the hash from the build, we
have to believe that it's the same thing. What's worse, I can't see
where in coho that we delete the hash from the build label.
I know that this was cited as one of the things that I was doing wrong
with the release process, but I have no idea why it's wrong to have
the hash in the header of the JS, since this is what you get when
manually generate the JS from the tag that is on the CordovaJS
repository. I think that this process isn't transparent, and I can't
find anywhere in the coho command that messes with this.
Anyone know why one is correct, and one is wrong? This seems pretty subjective.
Joe
Re: Cordova JS, CordovaWebView and Coho
Posted by Andrew Grieve <ag...@chromium.org>.
On Wed, Sep 18, 2013 at 3:10 PM, Joe Bowser <bo...@gmail.com> wrote:
> OK, that's pretty damn dishonest. I did pull and see that it was
> retagged like you said it was yesterday. I think blaming me for using
> the old tag back before you retagged is a pretty crap thing to do.
>
Woah - I'm not blaming here, I don't actually even know for sure what's
going on. I just suggested that the symptoms you describe (about grunt
producing a header with a commit hash) is the reason I re-tagged the JS
(which I did yesterday), so it is likely explained by not pulling since
then. Not accusing you of anything here, just trying to help.
>
> Also, Why in the hell are we storing the version in
> CordovaWebView.java? Does it need to be there? I thought that we've
> gone past having to hardcode Android versions in Java files.
>
https://issues.apache.org/jira/browse/CB-4725 - We have always had the
version hardcoded in Java. It was broken in 3.0 though by the Device plugin
being moved into a plugin. I put it in CordovaWebView.java so that we can
fix the Device plugin.
>
> On Wed, Sep 18, 2013 at 12:04 PM, Andrew Grieve <ag...@chromium.org>
> wrote:
> > The extra hash on the end was the reason for the re-tag of cordova-js.
> Maybe
> > you forgot to "git pull" and still have your cordova-js at the previous
> tag?
> >
> > Coho's not involved in any of that. The code is in
> > cordova-js/build/packager.js
> >
> >
> > On Wed, Sep 18, 2013 at 2:35 PM, Joe Bowser <bo...@gmail.com> wrote:
> >>
> >> After I let Andrew do the tagging of RC1, I noticed something that
> >> looks broken by the fact that I can't reproduce this result without
> >> using coho, and I can't find in the source where coho messes with the
> >> build labels.
> >>
> >> Now, as well all know, the JS is generated by Grunt. Assuming that
> >> we're going to be building off the same branch for the JS, we should
> >> all be getting the same JS by doing this:
> >>
> >> git checkout 3.1.0-rc1
> >> grunt
> >>
> >> That produces a JS file with this header:
> >> 3.1.0-rc1-0-g0d70465
> >>
> >> However, when you look at the JS checked into Android, it's simply just
> >> this:
> >> 3.1.0-rc1
> >>
> >> Now, they're the same, but when we remove the hash from the build, we
> >> have to believe that it's the same thing. What's worse, I can't see
> >> where in coho that we delete the hash from the build label.
> >>
> >> I know that this was cited as one of the things that I was doing wrong
> >> with the release process, but I have no idea why it's wrong to have
> >> the hash in the header of the JS, since this is what you get when
> >> manually generate the JS from the tag that is on the CordovaJS
> >> repository. I think that this process isn't transparent, and I can't
> >> find anywhere in the coho command that messes with this.
> >>
> >> Anyone know why one is correct, and one is wrong? This seems pretty
> >> subjective.
> >>
> >> Joe
> >
> >
>
Re: Cordova JS, CordovaWebView and Coho
Posted by Simon MacDonald <si...@gmail.com>.
http://static.giantbomb.com/uploads/original/4/40057/2482831-archer-1-phrasing.jpg
Simon Mac Donald
http://hi.im/simonmacdonald
On Wed, Sep 18, 2013 at 3:10 PM, Joe Bowser <bo...@gmail.com> wrote:
> OK, that's pretty damn dishonest. I did pull and see that it was
> retagged like you said it was yesterday. I think blaming me for using
> the old tag back before you retagged is a pretty crap thing to do.
>
> Also, Why in the hell are we storing the version in
> CordovaWebView.java? Does it need to be there? I thought that we've
> gone past having to hardcode Android versions in Java files.
>
> On Wed, Sep 18, 2013 at 12:04 PM, Andrew Grieve <ag...@chromium.org> wrote:
>> The extra hash on the end was the reason for the re-tag of cordova-js. Maybe
>> you forgot to "git pull" and still have your cordova-js at the previous tag?
>>
>> Coho's not involved in any of that. The code is in
>> cordova-js/build/packager.js
>>
>>
>> On Wed, Sep 18, 2013 at 2:35 PM, Joe Bowser <bo...@gmail.com> wrote:
>>>
>>> After I let Andrew do the tagging of RC1, I noticed something that
>>> looks broken by the fact that I can't reproduce this result without
>>> using coho, and I can't find in the source where coho messes with the
>>> build labels.
>>>
>>> Now, as well all know, the JS is generated by Grunt. Assuming that
>>> we're going to be building off the same branch for the JS, we should
>>> all be getting the same JS by doing this:
>>>
>>> git checkout 3.1.0-rc1
>>> grunt
>>>
>>> That produces a JS file with this header:
>>> 3.1.0-rc1-0-g0d70465
>>>
>>> However, when you look at the JS checked into Android, it's simply just
>>> this:
>>> 3.1.0-rc1
>>>
>>> Now, they're the same, but when we remove the hash from the build, we
>>> have to believe that it's the same thing. What's worse, I can't see
>>> where in coho that we delete the hash from the build label.
>>>
>>> I know that this was cited as one of the things that I was doing wrong
>>> with the release process, but I have no idea why it's wrong to have
>>> the hash in the header of the JS, since this is what you get when
>>> manually generate the JS from the tag that is on the CordovaJS
>>> repository. I think that this process isn't transparent, and I can't
>>> find anywhere in the coho command that messes with this.
>>>
>>> Anyone know why one is correct, and one is wrong? This seems pretty
>>> subjective.
>>>
>>> Joe
>>
>>
Re: Cordova JS, CordovaWebView and Coho
Posted by Joe Bowser <bo...@gmail.com>.
OK, that's pretty damn dishonest. I did pull and see that it was
retagged like you said it was yesterday. I think blaming me for using
the old tag back before you retagged is a pretty crap thing to do.
Also, Why in the hell are we storing the version in
CordovaWebView.java? Does it need to be there? I thought that we've
gone past having to hardcode Android versions in Java files.
On Wed, Sep 18, 2013 at 12:04 PM, Andrew Grieve <ag...@chromium.org> wrote:
> The extra hash on the end was the reason for the re-tag of cordova-js. Maybe
> you forgot to "git pull" and still have your cordova-js at the previous tag?
>
> Coho's not involved in any of that. The code is in
> cordova-js/build/packager.js
>
>
> On Wed, Sep 18, 2013 at 2:35 PM, Joe Bowser <bo...@gmail.com> wrote:
>>
>> After I let Andrew do the tagging of RC1, I noticed something that
>> looks broken by the fact that I can't reproduce this result without
>> using coho, and I can't find in the source where coho messes with the
>> build labels.
>>
>> Now, as well all know, the JS is generated by Grunt. Assuming that
>> we're going to be building off the same branch for the JS, we should
>> all be getting the same JS by doing this:
>>
>> git checkout 3.1.0-rc1
>> grunt
>>
>> That produces a JS file with this header:
>> 3.1.0-rc1-0-g0d70465
>>
>> However, when you look at the JS checked into Android, it's simply just
>> this:
>> 3.1.0-rc1
>>
>> Now, they're the same, but when we remove the hash from the build, we
>> have to believe that it's the same thing. What's worse, I can't see
>> where in coho that we delete the hash from the build label.
>>
>> I know that this was cited as one of the things that I was doing wrong
>> with the release process, but I have no idea why it's wrong to have
>> the hash in the header of the JS, since this is what you get when
>> manually generate the JS from the tag that is on the CordovaJS
>> repository. I think that this process isn't transparent, and I can't
>> find anywhere in the coho command that messes with this.
>>
>> Anyone know why one is correct, and one is wrong? This seems pretty
>> subjective.
>>
>> Joe
>
>
Re: Cordova JS, CordovaWebView and Coho
Posted by Andrew Grieve <ag...@chromium.org>.
The extra hash on the end was the reason for the re-tag of cordova-js.
Maybe you forgot to "git pull" and still have your cordova-js at the
previous tag?
Coho's not involved in any of that. The code is in
cordova-js/build/packager.js
On Wed, Sep 18, 2013 at 2:35 PM, Joe Bowser <bo...@gmail.com> wrote:
> After I let Andrew do the tagging of RC1, I noticed something that
> looks broken by the fact that I can't reproduce this result without
> using coho, and I can't find in the source where coho messes with the
> build labels.
>
> Now, as well all know, the JS is generated by Grunt. Assuming that
> we're going to be building off the same branch for the JS, we should
> all be getting the same JS by doing this:
>
> git checkout 3.1.0-rc1
> grunt
>
> That produces a JS file with this header:
> 3.1.0-rc1-0-g0d70465
>
> However, when you look at the JS checked into Android, it's simply just
> this:
> 3.1.0-rc1
>
> Now, they're the same, but when we remove the hash from the build, we
> have to believe that it's the same thing. What's worse, I can't see
> where in coho that we delete the hash from the build label.
>
> I know that this was cited as one of the things that I was doing wrong
> with the release process, but I have no idea why it's wrong to have
> the hash in the header of the JS, since this is what you get when
> manually generate the JS from the tag that is on the CordovaJS
> repository. I think that this process isn't transparent, and I can't
> find anywhere in the coho command that messes with this.
>
> Anyone know why one is correct, and one is wrong? This seems pretty
> subjective.
>
> Joe
>