On Fri, Sep 25, 2009 at 4:46 PM, <[email protected]> wrote:

>
> http://gwt-code-reviews.appspot.com/72802/diff/1/5
> File dev/core/src/com/google/gwt/dev/About.java (right):
>
> http://gwt-code-reviews.appspot.com/72802/diff/1/5#newcode50
> Line 50: * @deprecated use {...@link #getGwtVersionObject()} or
> On 2009/09/25 23:25:39, Ray Ryan wrote:
>
>> I take it getGwtVersionArray was not public api? Would library writers
>>
> be
>
>> relying on it?
>>
>
> The point is even if they were it cannot represent these version
> numbers.  Seems better to break them so they use the correct API rather
> than silently get incorrect results.
>
> http://gwt-code-reviews.appspot.com/72802/diff/1/5#newcode97
> Line 97: gwtVersionNum = tmp;
> On 2009/09/25 23:25:39, Ray Ryan wrote:
>
>> Seems like gwtVersionNum field should go away, and String
>>
> getGwtVersionNum()
>
>> should now be gwtVersion().getNumberString(), something like that.
>>
>
> That would be more of a breaking change.
>
>  For that matter, I wonder if GwtVersion should take over the gwtName
>>
> as well? So
>
>> its constructor would actually be GwtVersion(String name, String
>>
> number)? (Less
>
>> certain of this, please push back.)
>>
>
> Then it isn't just a GwtVersion, but a ProductInfo or something similar.
>  I'm ok with that, but that seems like feature creep.
>
> http://gwt-code-reviews.appspot.com/72802/diff/1/7
> File dev/core/src/com/google/gwt/dev/GwtVersion.java (right):
>
> http://gwt-code-reviews.appspot.com/72802/diff/1/7#newcode135
> Line 135: public enum ReleaseType {
> On 2009/09/25 23:25:39, Ray Ryan wrote:
>
>> why public
>>
>
> If GwtVersion is the public API for getting a structured version number,
> you need to be able to make sense of its components.


>
> http://gwt-code-reviews.appspot.com/72802/diff/1/7#newcode187
> Line 187: * soley of digits, it will be stored as an integer component
> instead.
> On 2009/09/25 23:25:39, Ray Ryan wrote:
>
>> Why?
>>
>
> For proper sorting, so 2 comes before 12.
>
> http://gwt-code-reviews.appspot.com/72802/diff/1/7#newcode401
> Line 401: public List<VersionComponent> getComponents() {
> On 2009/09/25 23:25:39, Ray Ryan wrote:
>
>> I don't see any calls to this. If I didn't miss some, why is it
>>
> public? Why does
>
>> it exist?
>>
>
> As above, if this is to be the recommendation for how external code
> deals with GWT versions, it needs to be usable as such.


To me this is the fundamental question: how did we get from nicer version
strings to a public api for dealing with GWT versions? Who is asking for
this new feature?

Anyway, I think Andrew has made the call that this is not an MS1 issue, so
this can go back on ice for now.

>
>
> http://gwt-code-reviews.appspot.com/72802/diff/1/3
> File dev/core/test/com/google/gwt/dev/AboutTest.java (right):
>
> http://gwt-code-reviews.appspot.com/72802/diff/1/3#newcode55
> Line 55: public void tstGwtVersionObject() {
> On 2009/09/25 23:25:39, Ray Ryan wrote:
>
>> bad test name. This isn't running.
>>
>
> Good catch.
>
> http://gwt-code-reviews.appspot.com/72802/diff/1/2
> File dev/core/test/com/google/gwt/dev/GwtVersionTest.java (right):
>
> http://gwt-code-reviews.appspot.com/72802/diff/1/2#newcode100
> Line 100: * Test that GwtVersion.compareTo produced expected results.
> On 2009/09/25 23:25:39, Ray Ryan wrote:
>
>> s/compareTo/equals/
>>
>
>  This is why I think javadoc on test methods is stupid. But I digress.
>>
>
> Ok.
>
> http://gwt-code-reviews.appspot.com/72802/diff/1/2#newcode197
> Line 197: assertTrue(components.get(1).isInteger());
> On 2009/09/25 23:25:39, Ray Ryan wrote:
>
>> Seems like a lot of this kind of thing should be in a separate unit
>>
> test on
>
>> VersionComponent. You could get rid of a lot of these redundant
>>
> isInteger checks
>
>> that way w/o reducing coverage.
>>
>
> Ok.
>
> http://gwt-code-reviews.appspot.com/72802/diff/1/4
> File dev/core/test/com/google/gwt/dev/shell/CheckForUpdatesTest.java
> (left):
>
> http://gwt-code-reviews.appspot.com/72802/diff/1/4#oldcode6
> Line 6: public class CheckForUpdatesTest extends TestCase {
> On 2009/09/25 23:25:39, Ray Ryan wrote:
>
>> Seems like you still need at least some kind of test at this level, to
>>
> make sure
>
>> that CheckForUpdates is actually using GwtVerison objects properly.
>>
>
>  It'd be silly for it to duplicate the coverage of GwtVersionTest, but
>>
> we still
>
>> at least want to see that isServerVersionNewer works at all.
>>
>
>  Also, how did you confirm that the old code will see the new style
>>
> strings as
>
>> newer?
>>
>
> It could be done by creating a fake server, but it is complicated.
> Since we didn't have it currently, it seems separate from this change.
>
> I'm not sure how to test old code short of copying it into the new code,
> but it just grabs exactly three integers and strips leading/trailing
> garbage.  I think that shouldn't be a problem since any of the new
> version numbers should be newer than the old numbers, but in any case
> there isn't anything we can do about the old code at this point.
>
> http://gwt-code-reviews.appspot.com/72802/diff/1/4#oldcode40
> Line 40: //    assertFalse(CheckForUpdates.isServerVersionNewer("2",
> "1.2.3"));
> On 2009/09/25 23:25:39, Ray Ryan wrote:
>
>> Oh. You're telling me that the tests I'm saying we shouldn't delete
>>
> have
>
>> actually never existed? Why is CheckForUpdates#isServerVersionNewer()
>> untestable?
>>
>
> isServerVersionNewer was simply a comparator between version numbers,
> and was obviated by the tests on GwtVersion.
>
>
> http://gwt-code-reviews.appspot.com/72802
>

--~--~---------~--~----~------------~-------~--~----~
http://groups.google.com/group/Google-Web-Toolkit-Contributors
-~----------~----~----~----~------~----~------~--~---

Reply via email to