On 2015-10-12, Scott Kostyshak wrote:
> On Mon, Sep 14, 2015 at 09:45:05AM +0000, Guenter Milde wrote:
>> ...

>> > 2) Unit test does not pass, investigation shows that the unit test is not 
>> > correct here and needs to be changed => code change + unit test change can 
>> > be submitted

>>   2a) unit test case was buggy,
>>   2b) unit test was correct but new behaviour is also correct.
...
>>   4) Unit test does not pass, investigation shows that the unit test is
>>   not correct here and needs to be changed => correct unit test reveals a
>>   different problem that was hidden by the incorrect test case and fails
>>   without further changes to the code.

>> ...

>> > For the case we are discussing it is already too late to submit the
>> > test change at the same time as the source change, but it is not too
>> > late to verify that the fix for #9764 does not introduce a
>> > regression _before_ submitting.

>> ...

>> > The assumption of the tex2lyx tests is that only correct .lyx.lyx files
>> > are submitted to the repository. Obviouly, this assumption is not true
>> > for the current version of test-insets.lyx.lyx, but will be fixed by
>> > updating it.

>> However, as both, the current state of the code as well as the
>> "pre-commit 268bd00" state produce/require a wrong .lyx.lyx file, we
>> would either need to submit a changed but still wrong test case or submit
>> a state where the test still fails.

>> > Please let us revisit the tex2lyx changes separately (after the
>> > tests pass again). Both 268bd00 and the patch from #9764 do only
>> > concern lib/unicodesymbols, bugs in tex2lyx are a separate issue and
>> > should be treated as such.

>> Both, unicodesymbols as well as tex2lyx use(d) \b instead of
>> \textsubbar for the COMBINING MINUS SIGN BELOW (IPA: retracted or
>> backed articulation). The unit test cases test-insets.tex and
>> test-insets.lyx.lyx connect these issues. This is why I'd prefer to
>> treat them in one commit.

>> If you go a different way to reach a correct end-state, I am fine with
>> it.

...

>> Is there a pointer somewhere how to compile to a build dir?
>> I currently just run "make" in the repository base dir (as suggested in
>> lyx/INSTALL).


> Is there still discussion to be had here as far as fixing the broken
> tex2lyx tests or is it clear what must be done but we must first resolve
> Günter's subscription to lyx-cvs issue [1] so he can commit?

I plan to provide a patch for 
+ lib/unicodesymbols,
+ tex2lyx
+ concerned test files

Either the "soft release lock" is lifted somehow or somone else needs to
apply it...

BTW: I have some more "easyfix-patches" in the pipeline and on track.

Thanks,

Günter

> [1] http://www.lyx.org/trac/ticket/9792#comment:4


Reply via email to