[cairo-commit] test/Makefile.am
Carl Worth
cworth at kemper.freedesktop.org
Thu Apr 12 12:47:33 PDT 2007
test/Makefile.am | 62 ++++++++++++++++++++++++++++++++++++++-----------------
1 files changed, 43 insertions(+), 19 deletions(-)
New commits:
diff-tree 29adf591b7026092a809e9a27878b5b4e35051fb (from c5b2f9d5c22c2259a11f0cbfbca8128cd3ac5794)
Author: Carl Worth <cworth at cworth.org>
Date: Thu Apr 12 12:47:27 2007 -0700
move show-glyphs-many to the DISABLED_TESTS list
And add a detailed comment explaining the reasons for disabling each
test in the list.
diff --git a/test/Makefile.am b/test/Makefile.am
index 3056969..7ee8e4c 100644
--- a/test/Makefile.am
+++ b/test/Makefile.am
@@ -113,9 +113,50 @@ rel-path \
push-group \
zero-alpha
+# XXX: Here are some existing tests that are currently disabled for
+# one reason or another. They can still be built and run (manually)
+# but we don't currently run them as part of 'make check' to avoid
+# inflicting some particularly nasty problems on the user of the test
+# suite (described below).
+#
+# It would definitely be worthwhile to address these problems so that
+# these tests can get back into the test suite proper, (since the
+# problems do indicated bugs somewhere and we're less likely to get
+# those bugs fixed without the attention due to having these tests in
+# the suite).
+#
+# The reasons for disabling the tests are as follows:
+#
+# extend-reflect - Triggers an infinite (or near-infinite) bug in some
+# X servers in some circumstances. Details and cause unknown.
+#
+# show-glyphs-many - This stress test was exercising a particular bug
+# in the xlib surface code (exceeding the X11 protocol request
+# limit) when rendering several thousand glyphs at once. The
+# original xlib-surface bug is fixed now, but the test continues
+# to stress some other aspects of the test suite. For example,
+# when doing text as paths, the resuilting PostScript file is one
+# giant path that ghostscript has a particularly hard time
+# with. I'm disabling this test for now, since I don't care about
+# that performance problem in ghostscript. (But, there is a
+# similar performance problem when using cairo to rasterize the
+# equivalen giant path---from an SVG files say---so this might be
+# a useful kind of test to bring back again for performance (not
+# correctness) testing.
+# text-glyph-range - This test triggers the following assertion in cairo:
+#
+# lt-text-glyph-range: cairo-scaled-font-subsets.c:350:
+# _cairo_sub_font_collect: Assertion `collection->num_glyphs ==
+# collection->max_glyph + 1' failed.
+#
+# And as a result causes a failure to be recorded by the test
+# suite, (in spite of the XFAIL status of the test). So maybe
+# that's just a bug in the test rig that should just consider
+# the abort an XFAIL like any other.
DISABLED_TESTS = \
-text-glyph-range \
-extend-reflect
+extend-reflect \
+show-glyphs-many \
+text-glyph-range
# Then we have a collection of tests that are only run if certain
# features are compiled into cairo
@@ -159,23 +200,6 @@ endif
endif
endif
-# XXX: Here are some existing tests that are currently disabled for
-# one reason or another.
-#
-# show-glyphs-many - this stress test was exercising a particular bug
-# in the xlib surface code (exceeding the X11 protocol request
-# limit) when rendering several thousand glyphs at once. The
-# original xlib-surface bug is fixed now, but the test continues
-# to stress some other aspects of the test suite. For example,
-# when doing text as paths, the resuilting PostScript file is one
-# giant path that ghostscript has a particularly hard time
-# with. I'm disabling this test for now, since I don't care about
-# that performance problem in ghostscript. (But, there is a
-# similar performance problem when using cairo to rasterize the
-# equivalen giant path---from an SBG files say---so this might be
-# a useful kind of test to bring back again for performance (not
-# correctness) testing.
-
# All tests which have a reference image go here.
REFERENCE_IMAGES = \
a8-mask-ref.png \
More information about the cairo-commit
mailing list