Re: Octave 3.8.1-rc4 release candidate available for ftp

[PhilipNienhuis]

John W. Eaton wrote
> The fourth release candidate of Octave 3.8.1 is available now
> from alpha.gnu.org in the directory /gnu/octave:
> 
>    cecbbd9bed2182246e3961ae30b89ccf  octave-3.8.1-rc4.tar.gz  (md5sum)
> 
>    -rw-r--r-- 1 3003 3002 20688476 Feb 26 20:26 octave-3.8.1-rc4.tar.gz
> 
> Please help make the 3.8.1 release a success by building from this
> release candidate and reporting any problems.
> 
> Unless there are significant problems, this will become 3.8.2 within a 
> week or so.
> 
> 3.8.1-rc4 will soon be the default build for mxe-octave.

Cross-built MXE on Windows XP, run in the GUI as well in the CLI
PASS     11480
FAIL         0
XFAIL        9
SKIPPED     12

Building packages works OK.
Note: I built it with mxe-octave 05c91ad85d01 from over a month or more ago.
I tried to cross-build with a newer version as well, today updated to
22b42ac3fd68+, but that failed while building native-gcc.
I think I'ĺl clone the repo again and start anew.

I did find some peculiarities with 3.8.1-rc4:

- The GUI command (terminal) pane does have focus after starting (it accepts
input), but no cursor is shown. The cursor only shows up after first
clicking in another pane and then in the terminal pane.
A 3.8.0 Octave, compiled with the essentially same mxe-octave version,
immediately shows the cursor after starting.

- Initially I got 2 fails in ov-cell.cc, because while running the
__run_test_suite__ script the iscomplex.m functions from the (loaded)
linear-algebra package had been picked up rather than those in core octave. 
Intruiging as 'which iscomplex' gave the functions in core octave
(libinterp/corefcn/data.cc).

-  After restarting Octave and running __run_test_suite__ as first command,
all tests pass.
However, any second run (and in fact, any subsequent run) of
__run_test_suite__.m, immediately after the first run, gives (GUI and CLI):

PASS     11420
FAIL        60
XFAIL        2
SKIPPED     12

with 57 FAILs in assert.m and 3 FAILS in ctor-vs-method.tst. All fails in
assert.m yield something like "expected... but  got no error".
The fntests.log files ends after the test of __run_test_suite__.m, although
the script does complete in the terminal.

FYI, Octave-3.8.0 (built with the same mxe build chain) behaves a little
similar; first run of __run_test_suite__.m gives 0 fails, ensuing ones 4 (1
in bug-38691.tst, 3 in ctor-vs-method.tst). assert.m does pass all tests
repeatedly.

I'm not sure how a somewhat older mxe-octave build chain should have
anything to do with these observations. AFAIU if Octave gets built
successfully it should just pass all tests.

Philip



--
View this message in context: 
http://octave.1599824.n4.nabble.com/Octave-3-8-1-rc4-release-candidate-available-for-ftp-tp4662375p4662596.html
Sent from the Octave - Maintainers mailing list archive at Nabble.com.