On Thu, Jan 24, 2013 at 8:08 PM, Tommaso Cucinotta <cucino...@sssup.it> wrote:
> On 09/01/13 09:12, Scott Kostyshak wrote:
>>> There are however
>>> two other areas that could be tested quite easily:
>>>
>>> a) Export of .lyx documents (to .tex, .html, whatever)
>>> b) lyx2lyx
>>>
>>> In both cases you would have a set of input files, and compare the generated
>>> output with a reference. I bet that introducing those tests would find bugs
>
>> OK this is good to know and does seem easy and useful. I might look
>> into this in a few months but since I haven't worked with exporting
>> much I might not get to it for a while longer.
>
> There's a partial testing of lyx2whatever conversions: the 
> autotests/export-in.sh
> script tries to test conversion of all documentation in lib/doc/, to xhtml
> and "lyx16x" formats. Can't remember why just those, guess one of those 
> conversions
> was triggering a crash.

Kornel just added a few more formats, pdf, pdf2, and pdf5. This (when
I ran it locally) caught some failures in the math manual. These three
extra formats were only implemented for CMake though.

>
> So, if none of these conversions crash, the test succeeds, otherwise you can 
> enjoy
> seeing which formats and .lyx doc files have troubles.
> I don't know whether this alternate launch of .sh scripts has been also 
> ported to
> the recent cmake infrastructure.

Yes, Kornel ported it fully to export-in.sh in the same directory.

> For what it matters, one can trivially include in the tested
> formats all the supported LyX "-e XXX" formats. Just, it's going to take a 
> huge time
> to complete, so a bit boring to run... :-) -- a remedy might be to 
> multi-core-ify
> those lyx export requests, as they're all batch and none of them requires the 
> GUI.

I figure it's much easier to get rid of tests than to write them. From
what I understand, no one doesn't use the tests because they take so
long to run. I think there are other reasons.

I agree that multi-core-ifying the export tests would be nice.

Thanks for the responses,

Scott

Reply via email to