Bug 37164 - testtool complains Unknown Command for legal script when running bas.
Summary: testtool complains Unknown Command for legal script when running bas.
Alias: None
Product: LibreOffice
Classification: Unclassified
Component: BASIC (show other bugs)
(earliest affected)
3.4.0 Beta5
Hardware: Other All
: medium major
Assignee: Noel Power
Keywords: regression
Depends on:
Reported: 2011-05-13 01:00 UTC by Yifan Jiang
Modified: 2011-09-21 07:10 UTC (History)
1 user (show)

See Also:
Crash report or crash signature:

patch for 3.4 (25.82 KB, patch)
2011-09-20 01:45 UTC, Noel Power

Note You need to log in before you can comment on or make changes to this bug.
Description Yifan Jiang 2011-05-13 01:00:30 UTC
Platform: SLED 11 sp1 i586
build info: Libreoffice 3.4 beta5


    1. Install LibO 3.4 beta5 as well as testtool package

    2. Run f_basic_first.bas using:

        /opt/libreoffice3.4/basis-link/program/testtool.bin -run $TESTAUTOMATION/framework/optional/f_basic_first.bas

    When the testtool step through the ResetApplication in master.inc, it
    complains an error

        : Unknown command ''

[Extra Information]

    - I reproduced the problem in 2 platforms, so pretty sure it was something
      wrong there.

    - Regression problem against 3.3
Comment 1 Yifan Jiang 2011-05-13 01:03:37 UTC
Hi Noel, since this is constantly reproducible in a clean environment. I opened the bug for your review.  Thanks!

P.S It might also influence other 'internal command' of testautomation thus make the automation test not runnable.
Comment 2 Alex Thurgood 2011-05-13 01:23:50 UTC
Hmmm, I just left a comment on the dev mailing list about the smoketests failing during make check when building from master ? Something to do with failed assertions in officeconnection.cxx - linked ?

Comment 3 Noel Power 2011-09-20 01:45:58 UTC
Created attachment 51392 [details]
patch for 3.4

fixes ( or appears to ) fix the testtool problems on 3.4
I tested with f_topten.bas ( because it is a short test ) Note: I started running the one mentioned above but after some time it was clear this test was going to take too long so I tried f_topten.bas instead.
The root cause of the problems was cws changehid which replaced numerical hid's by strings. Additionally the class smarthid was removed.

The patch imo is not ideal, . It appears that there are still hid(s) that are stringified numbers ( or at least that's what the testtool code expects ) and now the testtool code goes to extra trouble to convert between strings/numbers. I would guess this makes the testtool slower. This should be changed on master, for 3.4 I think a "just get it working" approach might be the best
Comment 4 Noel Power 2011-09-21 07:10:43 UTC
fixed on 3.4 branch