#171 3.2.9 (and before) testsuite: 17 106 108 failed

[David Runge]

Hi! When updating the package for Arch Linux I noticed failing tests:

Embed postscript variants.

104: epsi                                            ok
105: postscript, created by fig2dev                  ok
106: postscript, created by latex                    FAILED (bitmaps.at:249)
107: find /MediaBox in pdf file                      ok
108: pdf                                             FAILED (bitmaps.at:260)

Currently I can only remove those tests to be able to update the package.

[David Runge]

Further test output can be found in attachment

[David Runge]

Interestingly enough, removing e.g. bitmaps.at or parts of it still runs those tests.
I'm at a loss how to disable or remove singular tests and will not run tests.

[David Runge]

Ah! Fooled by autotools yet again, haha.
If I remove the pre-generated fig2dev/tests/testsuite script my changes actually apply and by removing the relevant tests I can "disable" them.

Seeing that only patching seems to help (maybe due to lack of understanding how this works): Is there any easier way to disable specific tests?

[Roland Rosenfeld]

The test 106 (bitmaps.at:249) should be a known incompatibility with new ghostscript 10.02. I reported it in https://sourceforge.net/p/mcj/tickets/160/ and it is already fixed in GIT https://sourceforge.net/p/mcj/fig2dev/ci/ebbb9e22afa6b2646d8cd9c359d62b604fc7f9d4/

But I have no idea, what's going wrong in test 108 (bitmaps.at:260). This should simply convert a small fig file to pdf, but fails on your system with a ghostscript error. Since this issue doesn't happen on Debian systems (which also uses gs 10.02.1), deeper debugging on your system seems to be necessary.

[tkl]

The log to test 17 seems to indicate that a temporary file could not be read. The fig file under test contains an embedded pdf, but a pdf.gz is available. Fig2dev uncompresses that to a temporary .pdf, from which ghostscript should read. But gs does not find or is unable to read that file.
Fixing these gs errors in the testsuite seems like trying to hit a moving target. For the time being, I would not like to investigate this specific case further.
Excluding tests from being run could be done either when directly running ./testsuite, or by passing an environment variable to make check:

./testsuite 1-16 18-105 107
# or
make check TESTSUITEFLAGS="1-16 18-105 107"

[tkl]

Looking again, failure of test 17 might also be connected to the issue reported in ticket [#161]. I will close this ticket, assuming that these issues are fixed with the pending commits for the next fig2dev version.