Project won't finish linking

I’ve been working on converting a window and a number of classes from a desktop project to a web project. I finally got it down to “0 errors” when trying to build. Yay!

Now, for Linux, the build never finishes linking the 64-bit project. it just sits with the CPU pegged at 99%-100%.

I tried to run locally on my Mac and while Safari launched the application never came up.

I have a number of controls subclassed from a web canvases. I did a Console search while building for Linux for “xojo” and nothing appeared. I even cleaned out the caches & restarted Xojo.

Ideas on where to start hunting? Thanks.

How long have you let it wait?

That seems to have been the key, at least for Linux. It takes about 10 minutes. That’s nuts. When the Linux version launches, I immediately get an “unhanded nil object exception”.

I’ve put Msgbox calls in all the 1st page’s events to see what’s happening, if anything. Previously there were no events. So I should just see my window/page.

I’m still waiting for the Mac debug build to finish “launching”. At least that’s what Safari says.

Maybe I’ll have to keep uploading Linux versions.

Just tried launching the version with the Msgboxes on Linux and the first thing that appears is:

[quote]Unhandled NilObjectException


64 bit uses the LLVM linker which does a TON more than the old compiler ever did - so yes it takes longer
Quite a bit longer in fact

Yes. I just timed it again. 9:03.11 per Linux/64bit compile. I’m not sure if I’ve been waiting 10 minutes for a local test (Mac/32bit). The browser launches but I don’t see anything. Not even the nil object exception that I see on Linux. I just put some msgbox calls in the open events of some subclassed controls.

10 minutes to build is not that odd
The old compiler basically took input code & spit out byte code immediately
More or less not optimized in any way

LLVM however does do a fair bit of optimization
There have been benchmark tests between the old RB Script vs the new LLVM based Xojo script if you are interested in comparisons of “unoptimized” vs “optimized” code speed

That said there’s likely to be some ability to select low/medium/high optimizations or something.
Some user selectability but its not settled exactly what yet.
Dunno if there will be any choices that include optimizing solely for speed vs space or what.

But yes … at the moment building & linking for 4 bit takes quite a bit longer