Carlo, thank you for your input. As you can probably tell from my inclusion of debug output, which you noted, I did run this in a debugger. Is "release with debug info" not the correct solution configuration, or are you suggesting I use "debug", instead? The last time I tried that I ran into all kinds of trouble and was told to use "release with debug info". Has that changed?
If you wish to be helpful, perhaps you could suggest why the debugger put the statistics reporting call in the middle of my call stack in some microsoft DLL. That was my original question. Am I reading it wrong, or did I do something wrong to get it to report like that? It reported the crash in the statistics code, which turned out to be incorrect, as far as I can tell, and sent me looking in the wrong direction. That crash was actually happening in my code, which does seem to contain debug info since I was able to breakpoint and examine it, once I knew it was crashing there. Can you tell from my posted call stack that it was crashing in my code and not the statistics reporting code? It would be most helpful if you could tell me where I went wrong and why I was sent looking at the wrong code. As I explained in my postings, I thought the difference I was seeing was that the code I was running (derived from the most current official OS release) was doing some statistics collection that the release code was not, and so I wished to not do that reporting, since the debugger told me it was crashing there. There was statistics collection added to recent viewers, was there not? So I asked if there was a way to turn it off. Of course it's not a good idea to just delete code that is crashing, but did I not explain that the code that appeared to be crashing was code I didn't want, anyway? It did occur to me that it was a bit selfish to not chase what appeared to be a bug that I appeared able to reproduce. My apologies if that seemed bad of me. I did want to get the thing stable before I did more work. My apologies if that seemed impatient. As to running "patches", yes, I am building a custom viewer. I can think of no reason to use the OS code other than to build a custom viewer. Of course there is code in there that I wrote, which I suppose you could call "patches". Should I post all of that code before asking questions? Thank you, Anna On Mon, Sep 28, 2009 at 7:03 AM, Carlo Wood <[email protected]> wrote: > On Sun, Sep 27, 2009 at 03:00:35PM -0400, Anna Gulaev wrote: > > This turned out to be my name cache observer. LL has moved from c_strings > to > > strings in a number of places since I wrote it. > > If you report a bug, you should always use the (latest) source code of > the people you report the bug to, and not have applied patches that you > don't tell them about :/ > > Also, removing code because you crash there is never ever ever > the right thing. The correct thing would have been to compile > the viewer yourself with debug support (no optimization and with > debugging symbols) and then run it inside a debugger; otherwise > any "backtrace" is often useless or even nonsense. > > Adding debug output, as you did, is a good technique too. > > -- > Carlo Wood <[email protected]> >
_______________________________________________ Policies and (un)subscribe information available here: http://wiki.secondlife.com/wiki/SLDev Please read the policies before posting to keep unmoderated posting privileges
