> On Nov. 9, 2012, 2:05 p.m., Richard Nelson wrote: > > indra/newview/llappviewer.cpp, line 3376 > > <http://codereview.secondlife.com/r/607/diff/1/?file=8087#file8087line3376> > > > > Structurally, it seems like this would be cleaner if opening the file, > > writing out the marke version, and closing it all happened in one function. > > Otherwise, you have this constraint that the file needs to be opened > > before being passed into recordMarkerVersion and closed afterwards. > > > > Just picking a nit, sorry.
I didn't do that because the usage of the two marker files is so different: the exec marker file is locked, but the logout marker file is not; and there is at least one code path in which the logout marker file is (apparently deliberately) not closed at all. - Oz ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: http://codereview.secondlife.com/r/607/#review1273 ----------------------------------------------------------- On Nov. 2, 2012, 1:55 p.m., Oz Linden wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > http://codereview.secondlife.com/r/607/ > ----------------------------------------------------------- > > (Updated Nov. 2, 2012, 1:55 p.m.) > > > Review request for Viewer and Callum Prentice. > > > Description > ------- > > In all the marker files used to detect how the viewer run terminates, record > the version. When checking the results, report errors only if the current > version matches the version in the file. This prevents errors in one version > from being reported against the subsequent version. > > > This addresses bug storm-1850. > http://jira.secondlife.com/browse/storm-1850 > > > Diffs > ----- > > indra/newview/llappviewer.h 3d35a13561fc > indra/newview/llappviewer.cpp 3d35a13561fc > > Diff: http://codereview.secondlife.com/r/607/diff/ > > > Testing > ------- > > Several simulated crashes both of the modified and unmodified viewers, and > some in which the marker file was modified manually to simulate different > viewers. Launched the new viewer after different crashes (and normal exits) > and confirmed (using logging temporarily added for that purpose) that the > reported last exec event was correct - and is always reported as Normal if > the previous version and the running version were not the same. > > > Thanks, > > Oz Linden > >
_______________________________________________ Policies and (un)subscribe information available here: http://wiki.secondlife.com/wiki/OpenSource-Dev Please read the policies before posting to keep unmoderated posting privileges
