Tue 02 Jun 2015 04:37:59 PM UTC, comment #12:
Retagging release from 4.0.0-rc1 to 4.0.0.
|
Fri 20 Mar 2015 03:07:30 PM UTC, comment #11:
I believe the remaining 3 failures are now fixed in the development code. jwe will be releasing a new RC2 candidate shortly. If there are remaining bugs they can be filed against that version.
Marking report fixed and closing.
|
Wed 11 Mar 2015 05:45:37 PM UTC, comment #10:
summary of failures in revision http://hg.savannah.gnu.org/hgweb/octave/rev/0cc365dada52 with 32bit build:
3 failures:
unpack.m
complex.tst
fail.m
the failure of fail is a new one
test tar no longer changes path
|
Sun 08 Mar 2015 01:37:24 AM UTC, comment #9:
The variables.cc tests should pass after this cset (http://hg.savannah.gnu.org/hgweb/octave/rev/85c5a81c6ddd).
The osmesa tests should pass after this cset (http://hg.savannah.gnu.org/hgweb/octave/rev/d9f6260cf20a).
The "issue" or "bug" with Octave not being able to close when there is a terminal running a program should be reported as another item. It won't be seen by anyone else since the title of the bug report is about test failures in _run_test_suite_.
|
Sat 07 Mar 2015 10:16:32 PM UTC, comment #8:
For the variables, test, If I start in a directory that doesn't have folders, the test passes the first time.
|
Sat 07 Mar 2015 09:56:46 PM UTC, comment #7:
Scratch inputParser fail off the list of failures.
It was grabbing file I had my work directory,
|
Fri 06 Mar 2015 05:00:47 PM UTC, comment #6:
The exit might be fixed via http://hg.savannah.gnu.org/hgweb/octave/rev/7d888f9e94eb ??
But if not, it is a bug :)
I would call the 'main window' the thing that determines whether octave should be closed or not - some programs may not require the terminal to run. However, another question then is, if the terminal is closed, and a program is run that uses the terminal, should it auto show itself?
|
Fri 06 Mar 2015 04:42:57 PM UTC, comment #5:
@comment #3:
Yep it is definitely https://savannah.gnu.org/bugs/?41699
Just:
test imread
exit
is enough.
OK, one issue (hmmm ...not solved, but...) identified.
Another thing:
While _run_test_suite_ is running in the GUI, selecting from the main menu bar:
File | exit
does nothing until _run_test_suite_ is finished - Octave shuts down then.
Should we consider this a bug? i.e., should "File | Exit" have a higher priority than anything running in Octave?
Similar goes to clicking the close button on the terminal pane. Is an octave w/o terminal still a "complete" Octave? Should the terminal always be present and visible?
FYI, in Matlab r2015a all panes have a "close" command on the right-click menu, except for the terminal.
|
Fri 06 Mar 2015 12:25:41 PM UTC, comment #4:
About the crash, maybe it is related to bug 41699 bug 41699, a long standing bug. The "test imread/imwrite" may cause this.
|
Fri 06 Mar 2015 12:23:23 PM UTC, comment #3:
The crash on exit has been happening for some time now, and is probally this one:
https://savannah.gnu.org/bugs/?41699
|
Fri 06 Mar 2015 12:17:27 PM UTC, comment #2:
For inputParser, I see at the top of the console, before the welcome text:
For the variables test, adding a variable to get the current directory at the top of the variables tests, it is starting in my home directory which has many folders and files, and so fails as it expects the third file entry to be a file.
|
Fri 06 Mar 2015 11:33:28 AM UTC, comment #1:
(Tested on win7 with
- my own --enable-windows-64 cross-build
- the 32/bit build from the gnu ftp site'
For me, variables.cc only fails in the first _run_test_suite_ run, it passes in all subsequent runs. A bit suspicious, uninitialized variable?
lscov.m fails (--enable-windows-64)
assert.m fails (--enable-windows-64)
axis.m fails (--enable-windows-64)
complex.tst only fails in the 32/bit build.
Both Octave versions (32b build and my own --enable-windows-64 build) crash consistently after having run _run_test_suite_.
Upon Ctrl-D, first the GUI disappears neatly, all looks good, but then I get the Windows popup that the program crashed somewhere (see attached pic). Maybe it is again related to Octave trying to write its settings file to an inappropriate place? (see bug #40444) The connection with _run_test_suite_ isn“t obvious however....
I've copied the info from that popup below:
(file #33258)
|
Fri 06 Mar 2015 09:06:35 AM UTC, original submission:
Summary of failures of Octave 4.0.0.-rc1 in Win7, when
running __run_test_suite__ as reported by JohnD:
@ variables.cc-tst - fails only in __run_test_suite__, running "test variables.cc-tst" pass
@ _osmesa_print_.cc-tst bug 44338
@ inputParser.m (passed for me)
@ unpack.m bug 43979
@ complex.tst bug 44206
@ another problem (not failure): test tar changes working directory bug 44007
|