Failures handled inelegantly on Linux

[Philip]

When failing to load Atlas, sometimes the game flashes something onto the screen too fast to read, and on the console it says:

udbg_init: loading symbols from ./pyrogenesis_dbg.
posting setcontext with canvas (nil)
ogl.cpp(370): Assertion failed: "("called before OpenGL is ready for use") && 0"
ogl.cpp(188): Assertion failed: "("GL_VERSION invalid") && 0"
ogl.cpp(188): Assertion failed: "("GL_VERSION invalid") && 0"
ogl.cpp(188): Assertion failed: "("GL_VERSION invalid") && 0"
ogl.cpp(188): Assertion failed: "("GL_VERSION invalid") && 0"
ogl.cpp(188): Assertion failed: "("GL_VERSION invalid") && 0"
addr2line: 'atlas': No such file
addr2line: 'atlas': No such file
addr2line: 'atlas': No such file
addr2line: 'atlas': No such file
addr2line: 'atlas': No such file
addr2line: 'atlas': No such file
addr2line: 'atlas': No such file
addr2line: 'atlas': No such file
addr2line: 'atlas': No such file
addr2line: 'atlas': No such file
addr2line: 'atlas': No such file
addr2line: 'atlas': No such file
addr2line: 'atlas': No such file
addr2line: 'atlas': No such file
addr2line: 'atlas': No such file
Aborted

On other occasions, the error windows stay up: there's Atlas - Scenario Editor - untitled and Atlas Error (from wxLogError) and Debug report "atlas" and Debug report "atlas" <2> (both from wxDebugReportPreviewStd). Only the last window is interactive. Clicking OK/Cancel usually causes the process to terminate, but once it just hung.

It'd be nicer if something nicer happened, with less random vanishing and hanging.

[Gallaecio]

With 0 A.D. 0.11, I am getting:

• A console-like error message about not being able to load Atlas’s .so file (GNU/Linux) at the top of 0 A.D.’s window, which is only displayed the first time I try to run the editor.
• A dialog window exmplaining the tool could not be loaded or is not available.

While you might want to consider not to display the console-like line at the top once the game is not an alpha version anymore — at least when the editor was disabled at build time —, in my opinion, the current approach is already nice enough.

Not sure if things are the same way on Windows machines, though.