Of the four options suggested in the jira, this is a modified version of #2. Since the mime_types.xml file actually gets localized (it contains a bunch of user-visible text in addition to the mappings), I don't think renaming it in viewer_manifest.py will work.
Instead, I've created additional files (mime_types_linux.xml and mime_types_mac.xml) in the same location, and changed the code that loads the file in LLAppViewer::init() to look for different names per-platform.
I'm sure there are cleverer solutions (such as abusing the translation xml overlay system to create files that have deltas per platform), but those don't fit in the time-box allocated to this issue.
Renamed llviewerversion to llversioninfo, to avoid confusion with
llversionviewer in llcommon (llversion is already used by llwindow).
Created new LLVersionInfo class with the following methods:
static S32 getMajor();
static S32 getMinor();
static S32 getPatch();
static S32 getBuild();
static const std::string &getVersion();
static const std::string &getShortVersion();
static const std::string &getChannel();
All viewer code has been updated to use this API. Viewer code no
longer directly includes llversionviewer.h from llcommon.
This module now contains general viewer version routines, not just
access to the viewer build number, so I've renamed it from
llviewerbuild to llviewerversion.
Rather than copy/paste the same version string formatting pattern
again, I added new string-based version routines:
/// return the full viewer version as a string like "2.0.0.200030"
const std::string &llGetViewerVersion();
/// return the viewer version as a string like "2.0.0"
const std::string &llGetViewerShortVersion();
/// return the viewer build version as a string, e.g., "200130"
const std::string &llGetViewerBuildVersion();
Review #49
This change adjusts each media's priority based on whether the viewer is minimized (media priority becomes HIDDEN) or unfocused (media priority becomes LOW). However, due to the fact that updateMedia() was no longer being called when minimized, I moved its call out of LLViewerTextureList::updateImages() (a seemingly odd place anyway) and into its own idle callback.
EXT-2456 : Usability: Inventory > Clothing > context menu not intuitive
RightClickFetchInventoryObserver now works correctly to update menu
FetchInventoryDescendents now triggers notifyObservers properly
Fixed up gInventory.notifyObservers logic
--HG--
branch : avatar-pipeline
This addresses a failure mode Mani encountered when lleventhost.dll wasn't
built. testlangs.py couldn't talk to the newly-launched viewer -- but thinking
that was because it had crashed, it went ahead and launched a whole bunch
more. This fix avoids the possibility of rogue viewer processes on an
unattended test machine.
Added gInventory.notifyObservers to idle callbacks so it gets triggered without explicit notifyObservers synchronization.
Added more state tracking for attachments, wearables, and links of those types, so that they're marked as changed properly.
initialized notifications screen channels after LLViewerWindow instantiated
and disabled all old notifications widgets(LLNotifyBox, LLGroupNotifyBox,
LLAlertDialog);
--HG--
branch : product-engine
and reduce unnecessary dependencies and incredibuild wedging. Hopefully in
the long run this will reduce build time. Also cleaned up a lot of header
file usage to conform better to the coding standard.
reviewed by james and steve.
with an event API. In addition to the LLEventPump name on which to listen,
LLEventAPI accepts a documentation string for event API introspection.
Give every LLEventDispatcher::add() overload a new documentation string
parameter for event API introspection.
Convert every existing event API to new conventions, introducing suitable
documentation strings for the API and each of its operations.
Normally LL_ERRS in a developer build of the viewer pops up an OS message box
with the text of the LL_ERRS log message, requiring user intervention to
proceed. This isn't good for unattended testing, especially on Parabuild
machines. Running the viewer with '--set QAModeTermCode n' for some
0 < n < 256 will, on LL_ERRS, cause the viewer to terminate immediately with
_exit(n).
The bug was driven by the following sequence in llstartup.cpp:
if(reason_response == "update"
|| reason_response == "optional")
{
// used to resend status event still containing "update",
// erroneously instantiating a second forced-update
// LLAlertDialog
LLLoginInstance::getInstance()->disconnect();
// quit with an LLAlertDialog still on sModalStack used
// to result in LL_ERRS
LLAppViewer::instance()->forceQuit();
}
I hope to be able to introduce a test script to verify the fix. That
script would need to be able to call LLAppViewer::forceQuit() rather than
requestQuit(), which is already available via LLAppViewerListener.
At the same time, changed LLAppViewerListener to bind a functor to
retrieve an LLAppViewer instance (namely LLAppViewer::instance) rather
than an LLAppViewer*. Apparently the static instantiation of
LLAppViewerListener was calling LLAppViewer::instance() too early, before
things were ready, so the declaration was changed to pass NULL -- then
in each method, call LLAppViewer::instance() if the bound pointer is NULL.
Binding the LLAppViewer* is a Feathers tactic intended to avoid the need
to reference the singleton. Binding a functor still leaves it up to the
instantiating code to reference LLAppViewer::instance, while deferring
the actual call to that method.
James Cook