[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: commit 12fe78825798191ecb7e5a4ee3064679773ae1ab is broken

From: James Lowe
Subject: Re: commit 12fe78825798191ecb7e5a4ee3064679773ae1ab is broken
Date: Sat, 09 Jun 2018 14:58:15 +0100 (BST)


On Fri, 8 Jun 2018 19:13:23 +0200, Knut Petersen <address@hidden> wrote:

> Am 08.06.2018 um 18:20 schrieb James Lowe:
> > How come I did not see anything different with the reg test output at the 
> > time?
> Maybe you did not see the correct images? pngs / jpegs might be cached by the 
> browser.
> At least konqueror 5.0.97 on my openSuSE Tumbleweed system definitely has 
> this problem.
> We could append a timestamp to the image urls, e.g.
> href="input/regression/out-test-baseline/test-output-distance.png?20180608191147"

Sorry to keep on about this - I just want to understand, that is all.

I don't think the browser is caching images.

However I decided to do check this 'semi-manually' by doing the following

Checking out the commit before any of those three that Malte checed in and then 
I built LilyPond and simply ran

lilypond -png $LILYPOND_GIT/input/regression

This produced a full set of png files which I copied to a separate directory

Then I checked out *just* the next 3 commits that Malte did (only because they 
were all for the same issue) and built that version of LilyPond and then ran 
the same command to get a second set of png files that I then copied to a 
second separate directory.

Finally I ran a simple bash 1-liner to read the files from the first dir and 
compare them with files of the same name in the second dir, usig imagemagicks 
compare CLI and output the 'diffs' to a third dir

e.g. from within the first dir of the pngs created before Malte's checkin:

for f in * ; do compare $f '../malte/'$f -compose Src '../diff/'$f ; done

This produced the resultant diff files which only showed the differing pixels.

I then ran through the diff files visually and only 2 files showed diffs; 

test-output-distance (obviously) and 'rest-dot-positions' (again I believe this 
is also expected at the  moment) but not other files showed any diffs.

Thanks again for your consideration.


reply via email to

[Prev in Thread] Current Thread [Next in Thread]