Bug#895725: librsvg: Test failures on s390x

[Simon McVittie]

On Sun, 15 Apr 2018 at 16:16:48 +1000, Tim Lunn wrote:
> # Storing test result image at /tmp/91-nonzero-translation-diff.png
> not ok 26 /rsvg-test/reftests/bugs/91-nonzero-translation.svg
> FAIL: rsvg-test 26 /rsvg-test/reftests/bugs/91-nonzero-translation.svg

It's helpful if you quote how many pixels differed and how much, since
that tells us whether the difference is likely to be human-visible.

This one looks bad:

# Storing test result image at /tmp/91-nonzero-translation-out.png
# 29023 pixels differ (with maximum difference of 89) from reference image

# Storing test result image at /tmp/91-nonzero-translation-diff.png
not ok 26 /rsvg-test/reftests/bugs/91-nonzero-translation.svg
FAIL: rsvg-test 26 /rsvg-test/reftests/bugs/91-nonzero-translation.svg

> not ok 39 /rsvg-test/reftests/svg1.1/filters-color-02-b.svg
> FAIL: rsvg-test 39 /rsvg-test/reftests/svg1.1/filters-color-02-b.svg
> # Storing test result image at /tmp/filters-composite-02-b-out.png

This one looks very bad:

# Storing test result image at /tmp/filters-color-02-b-out.png
# 35120 pixels differ (with maximum difference of 255) from reference image

# Storing test result image at /tmp/filters-color-02-b-diff.png
not ok 39 /rsvg-test/reftests/svg1.1/filters-color-02-b.svg
FAIL: rsvg-test 39 /rsvg-test/reftests/svg1.1/filters-color-02-b.svg

> # Storing test result image at /tmp/pservers-grad-08-b-diff.png
> not ok 93 /rsvg-test/reftests/svg1.1/pservers-grad-08-b.svg
> FAIL: rsvg-test 93 /rsvg-test/reftests/svg1.1/pservers-grad-08-b.svg

This one looks OK (7 pixels differing by 2 levels of a colour channel
is not going to be something you can see), but I'm a little concerned
by the parsing error:

# librsvg-MESSAGE: CSS parsing error

# librsvg-MESSAGE: CSS unrecoverable error

# Storing test result image at /tmp/pservers-grad-08-b-out.png
# 7 pixels differ (with maximum difference of 2) from reference image

# Storing test result image at /tmp/pservers-grad-08-b-diff.png
not ok 93 /rsvg-test/reftests/svg1.1/pservers-grad-08-b.svg
FAIL: rsvg-test 93 /rsvg-test/reftests/svg1.1/pservers-grad-08-b.svg

To be able to debug these, someone will need to build librsvg on a
porterbox and fetch the -diff.png images for inspection. I'll kick off
a build on zelenka later today if nobody gets there first.

I'm very tempted to increase the maximum delta from 1 to 2 and ask
upstream to do the same.

    smcv