OK, when the vtests are updated let me know, then I'll try generating
some SVG files and we'll see if they're consistent across platforms. Is
the problem the rendering of the PNG files on different platforms, or
the MuseScore generating of the PNG file across diff platforms? You say
rendering, but that strikes me as odd - if you're doing a diff across
the binary files, only different bits in the files will cause red
flags. That would point to cross-platform generation, not rendering, as
the problem. And if PNGs render differently across OS or across
browsers or graphic apps, there's not much you can do about that.
With SVG, I might expect to see differences in the rendering across
browsers, though now that MuseScore doesn't use any fonts in the SVG
that will be minimized.
--
James
On 4/27/2016 7:25 AM, Lasconic wrote:
All of them. Also the vtests are not in a good shape currently because
of the big merge.
Some are not even rendered because current master can't open 1.X files
and the tests were made with 1.3.
Werner is fixing the tests as we speak.
See http://vtest.musescore.org/994090f1/vtest.html for the current state.
lasconic
2016-04-27 15:02 GMT+02:00 Sideways Skullfinger
<[email protected]
<mailto:[email protected]>>:
Unless someone knows the answer to your question, we'll have to
generate some SVG files and see how they look across diff
platforms. For me, that requires other people with the other
platforms installed, as I am only running Windows 7. Are there
particular PNGs that are problematic and that should be tested
first across platforms?
--
James
On 4/27/2016 2:50 AM, Lasconic wrote:
So in my previous comment, I was wondering if it would make sense
to add SVG or replace PNGs by SVGs in the mtest.
I thought that it would solve a recurring problem: pngs rendered
on different platforms looks different, and so when we "diff"
them pixel by pixels we don't get blank files but some pixels are
different. So the vtests are currently not automated, we need to
check them from time to time and see if there is a change. We
don't have any automated "red flag".
My hope was that diffing SVG (and so XML) would avoid this
problem and so we could automate the diff and raise a red flag if
we find a regression in the layout/drawing algorithm. I'm not
sure if it's true or not and if it's not, then we need to solve
the root cause and find why different OS gives different rendering...
lasconic
2016-04-26 22:17 GMT+02:00 Sideways Skullfinger
<[email protected]
<mailto:[email protected]>>:
So is the goal here simply to duplicate the vtests using
SVG? Or is there additional automation can be applied to the
existing process? If you're already automating the diff
reporting, what more does SVG actually offer you? Yes, it
would test both the PNG and SVG export processes, but
relative to automation: I don't see the difference between
running a diff between two PNG files vs. two text files. Am
I missing something?
If it's just duplicating the existing vtests to use SVG, that
shouldn't be too hard.
--
Sideways
On 4/26/2016 2:06 PM, Marc Sabatella wrote:
vtests themselves are all created manually - the source
files and the reference graphics. The tests are run via the
"gen" script in the vtest folder - that's what actually
generates the tests files and diffs them against the
reference. Go to vtest and type ./gen (you might need to
configure some things first so the script uses the right
version of MuseScore) and you'll see the tests being run; at
the end an HTML file is generated that you can inspect manually.
On Tue, Apr 26, 2016 at 1:36 PM Sideways Skullfinger
<[email protected]
<mailto:[email protected]>> wrote:
On 4/22/2016 12:01 PM, Lasconic wrote:
> We don't have any SVG tests suite, but it's a good
idea, we could
> modify the vtests to render SVG too (or instead of PNG?)
> (If you don't know what the vtests are, check the
vtests directory and
> http://vtest.musescore.org/index.html)
>
> Diffing SVG is easier so we could even use them for
regression testing.
I now understand this a bit better, and want to
understand it fully in
order to make it happen. I assume the goal would be to
add SVG exports
to the mtests, and leave the PNG vtests around, at least
for now. But I
know very little about how all this is set up.
I now know that the automated testing is setup in the
mtests folder. I
am assuming that the vtest files are generated
automatically in mtests
somewhere, and I would need to duplicate that or add an
SVG export of
the same scores. Correct?
Then there is the automated diff process and reporting
on differences
encountered. Is there an existing way that this is done
in MuseScore?
Those are my questions for now, I'm sure I'll have more
as I get into it...
--
Sideways
------------------------------------------------------------------------------
Find and fix application performance issues faster with Applications Manager
Applications Manager provides deep performance insights into multiple tiers of
your business applications. It resolves application problems quickly and
reduces your MTTR. Get your free trial!
https://ad.doubleclick.net/ddm/clk/302982198;130105516;z
_______________________________________________
Mscore-developer mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/mscore-developer
