Code Log

Commit Date  
[r10772] by cannam

* Fix to the superficial problem reported at (thanks Arnout) -- though presumably something is still wrong
there if shutdown() is being called from createDriver() at all

2009-08-27 11:24:15 Tree
[r10771] by cannam

* Fix to the superficial problem reported at (thanks Arnout) -- though presumably something is still wrong there if shutdown() is being called from createDriver() at all

2009-08-27 11:23:33 Tree
[r10770] by cannam

* update descriptions of trunk & branches

2009-08-27 11:03:21 Tree
[r10769] by msjulie

Some preliminary menu juggling.
Assigning F3 and Shift+F3 for note / rest insert

no new progress.

2009-08-26 20:11:31 Tree
[r10768] by cjnfryer

* Fixed the crash on file->new (AKA Ilan's crash)

A slot of the old RosegardenMainViewWidget was firing (followed
by the new widget if the crash was survived) during this action:
in RosegardenMainWindow::initView().

Solution is to delete the oldView widget.

2009-08-26 18:59:34 Tree
[r10767] by cannam

* Merge from devicepush branch. Not all the details are done yet, but it's
broadly right in principle.

Still to look at:
* Renaming a device in the device manager does not rename its instruments
* Importing a studio from a file into the current document does not work
* When a connection is made externally (using qjackctl etc), the instrument
connection status and device manager don't update (need to restore a small
part of the system update instruments logic)
* Not yet handled setting a connection for first device in document if no
connection configured after file load

2009-08-26 16:25:12 Tree
[r10766] by cannam

* Tidy up

2009-08-26 15:10:39 Tree
[r10765] by cjnfryer

* Replaced the ViewElementList m_selectedElements in ControlRulerWidget with a simple std::vector

Fixes crash on Matrix Editor close.
In my defence, I would like to point out that I inherited use of said inappropriate class.
My crime, if it can be considered as such, was to use one as a data member so that it can be used
to display selected elements correctly when creating a property ruler ...

2009-08-26 10:02:32 Tree
[r10764] by msjulie

Got basic images for note-inserter and rest-inserter -- file names in flux
Got basic funtionality of toolbar actions for direct switch to note_inserter
mode and rest_inserter mode.

All rough all in flux, but progress

2009-08-26 03:05:36 Tree
[r10763] by cjnfryer

Confused/out of my depth.

Jani's valgrind o/p suggests my ControlRulerWidget::m_selectedElements list is the culprit for this crash:
==15885== Invalid free() / delete / delete[]
==15885== at 0x4C260AD: operator delete(void*) (vg_replace_malloc.c:342)
==15885== by 0x4C4CBB: Rosegarden::ViewElement::~ViewElement() (ViewElement.cpp:36)
==15885== by 0x4C500C: Rosegarden::ViewElementList::~ViewElementList() (ViewElement.cpp:72)
==15885== by 0x793E53: Rosegarden::ControlRulerWidget::~ControlRulerWidget() (ControlRulerWidget.cpp:58)

Presumably, because QT is trying to delete the elements contained in the list but they're already deleted?

This commit clears m_selectedElements during the ControlRulerWidget destructor and fixes the crash on my PC at least.
It doesn't feel right.

What don't I understand?

Is this the QT auto-cleanup mechanism? Should it be trying to cleanup a list of pointers?
If so, how am I supposed to say "don't clean these up, they belong to another widget"?

2009-08-25 22:09:13 Tree
[r10762] by cjnfryer

* Added the WA_DeleteOnClose attribute to EditViewBase and some RG_DEBUG reports in destructors.
* Crashes reliably when you close the matrix editor but only if you've already drawn a note.

The crash can be delayed by commenting out delete m_scene in ~MatrixWidget.
RG will then crash next time you open the segment and try to draw a note.

Looks to me like it's not getting through QT's child deletion process as comments in ~MatrixTool
are not showing but do if you comment out the delete of m_scene in ~MatrixWidget.

I don't have the QT source on this PC so gdb can't help in the location of the crash.
Hopefully someone with that installed and valgrind (or whatever) can help locate the problem.

2009-08-25 18:15:12 Tree
[r10761] by msjulie

This branch for hashing out shortcuts and toolbar issues in the notation

Join the fun if you like.

2009-08-25 18:07:16 Tree
[r10760] by cannam

* clear devices sequencer-side when loading a new document

2009-08-25 16:35:39 Tree
[r10759] by cannam

* Also commit to stable the library fixes just committed to trunk

2009-08-24 16:28:45 Tree
[r10758] by cannam

* This change (from the Qt4 "devpush" branch) is a pure bugfix, and should
be in stable_1_7 as well. So here it is. The devpush branch will be
merged up soonish, so I won't bother copying this fix to trunk on its own.

2009-08-24 16:27:03 Tree
[r10757] by cannam

* Thanks to Luke Briner for pointing out that the correct RD-700SX device
file we had included in older versions has somehow been overwritten with
a copy of the RD-700GX file in our most recent releases. Restoring the
original SX definition. Also, the GX device file was not compressed (I
wonder if there's any connection?) -- compressing it. It isn't immediately
obvious to me from the svn log what went wrong here.

2009-08-24 16:25:39 Tree
[r10756] by cannam

* Fixes to device reconnect & rename

2009-08-24 15:33:32 Tree
[r10755] by cannam

* never connect to system port in plausible connection lookup

2009-08-24 13:38:56 Tree
[r10754] by cannam

* Another incremental update. Includes ACTUAL EXPLANATORY COMMENT

2009-08-24 13:26:28 Tree
[r10753] by cannam

* of course, we also want tooltips for those actions that don't have shortcuts

2009-08-24 12:30:21 Tree
[r10752] by cannam

* Use IconLoader for makeToolbarPixmap. This restores the little diddy
note icons that appear in the delay/quantize combos in the segment
parameter box.

2009-08-24 12:04:24 Tree
[r10751] by cannam

* Add tooltips for note entry buttons. I realise I don't actually know
whether to say "quarter rest" or "quarter-note rest" in Americanese.
Perhaps a native speaker could review this.

2009-08-24 11:51:00 Tree
[r10750] by cannam

* Attempt to fix normalizeRest problems when inserting controller events,
and generally tighten up the logic in that function.
Mostly, this means only considering note, text, and (existing) rest events
as candidates for the boundary for the range to be normalized. Also
remove an erroneous re-assignment to endTime after the proper endTime has
already been calculated.

2009-08-24 10:35:39 Tree
[r10749] by msjulie

make dummy button non-selectable (enabled == false).

2009-08-24 03:27:40 Tree
[r10748] by dmmcintyr

This branch has served out its purpose, and is now obsolete.

2009-08-24 03:08:47 Tree
Older >