Sun 03 Sep 2017 01:15:18 PM UTC, comment #12:
Closing this old report.
On 4.3.0+ the OF windows package works fine
|
Mon 14 Aug 2017 08:52:51 PM UTC, comment #11:
Pushed http://hg.octave.org/mxe-octave/rev/1933a9113e65
|
Sat 07 Nov 2015 05:59:41 PM UTC, comment #10:
That's what I figured.
Actually OF packages with binary modules should depend on mkoctfile (= Octave) but only in case of --enable-binary-packages.
I'll have a first go at this later on, if no one beats me to it.
|
Sat 07 Nov 2015 01:53:47 PM UTC, comment #9:
Yeah I do believe there is no rule in there to force a rebuild if the octave target changes.
The easiest way perhaps woruld be to make octave a dependcay for all octave packages.
|
Fri 06 Nov 2015 08:51:04 PM UTC, comment #8:
So yes, recompiling the windows package does the trick.
Does mxe-octave need a mechanism to recompile all packages if the Octave target proper has changed?
I suppose that as all OF packages with binary modules depend on Octave, such a rule would be helpful.
OTOH I know I'm bending the system as I continuously feed "make all dist" targets to mxe-octave into an existing tree.
I now wonder if it'll work w/o --enable-binary-packages flag. I have no mxe build tree with that configure setting.
Shall I close this bug report with "Invalid", or shall I change the title to s/th like "OF packages in mxe-octave should depend on Octave" ?
|
Fri 06 Nov 2015 07:42:08 PM UTC, comment #7:
Hmmm... digging around I just noted that the mxe-cross built usr/..../src/windows-1.2.1.tar.gz package in the mxe-octave build tree dated back from August, 15 (I tend to re-use existing mxe-octave trees for some months, hg pull-updating them every once and a while).
I just touch-ed all src/of-*,mk files and now have a new build running.
If this works out OK I suppose some make interdependencies are wrongly configured. Or, I'm simply too naive with undue expectations :-) (very well possible)
I'll let you know how it goes.
|
Fri 06 Nov 2015 05:16:44 PM UTC, comment #6:
This was using a new build using configure settings: $ ./configure --enable-octave=default --enable-binary-packages --enable-windows-64=yes --enable-devel-tools
So sounds like it should be the same as yours. Just 64bit OS, 32b index
|
Fri 06 Nov 2015 05:00:48 PM UTC, comment #5:
I use that --enable-binary-packages flag as well.
Intriguing that is doesn't work with 64b Windows/32b indexing but does with 64b indexing.
BTW were you able to confirm my observation? i.e., does not work with W64/32b indexing ?
|
Fri 06 Nov 2015 03:02:02 PM UTC, comment #4:
Note, I used the binary package created with --enable-binary-packages when compiling mxe-octave
|
Fri 06 Nov 2015 11:35:53 AM UTC, comment #3:
|
Thu 05 Nov 2015 10:46:42 PM UTC, comment #2:
With 20678:4b00afb5e9c3 it still doesn't work for X86_64 builds (just --enable-windows-64).
Have you tried if COM really works:
i.e.,
app = actxserver ("Excel.Application")
app.Quit()
delete (app)
?
(no guarantees that COM will work with MS-Office anyway for 64-bit indexing Octave)
|
Thu 05 Nov 2015 06:07:36 PM UTC, comment #1:
Using current tip (changeset 20677:0d5d8db55790), in Windows, build in 64 bit mode works for me.
|
Sat 17 Oct 2015 12:05:47 PM UTC, original submission:
The OF windows package has stopped working with recent Octave development versions.
where the following is expected:
Experimenting with older Octave binary Windows installers I have lying around shows that OF windows packages from the mxe-octave installers:
- work fine with Octave rev. f61c67865d9f (~7 Oct.)
- don't work anymore with Octave rev. 56333f6df823 (~10 Oct.)
(Those revisions and dates are of course just indicative - somewhere in between them the windows package stopped working. I merely include the revision name when renaming the mxe-octave binary installers.)
The OF windows-1.2.1 package from the newest Octave binary I have (rev. ab705b42cfd8, ~15 Oct) installs and runs fine in an Octave installation from 7 Oct.
So something inside Octave itself has changed. Could this be the error_state changes?
Item Group => "Incorrect result" as I think core Octave changes affecting OF packages aren't to be perceived as "regressions".
|