Recent compiler versions broken ?

Jim Hearne · Joined: 22 Dec 2003 Posts: 109 Location: West Sussex, UK

Is PCWH 3.226,7 or 8 broken ?
I've code that works fine complied with 3.225 that goes nutty with 3.228, i've not tried the in between versions yet.

Anybody else have problems with recent releases ?
I'm using a 16F876A and 3.228 is causing software counters to expire much sooner than they should be amongst other problems.
Recompile the same code with 3.225 and it's fine.

Jim

valemike · Guest

I have circuitry on my board that can blow a fuse or damage components if my code goes wrong. I'm compiling with 3.225 right now, but have all the later versions installed. That's why i wait for forum feedback, like your post, before i prematurely try it out on my board Idea

If it was a non-dangerous design, i wouldn't mind being a guinea pig for a beta version. Isn't 3.191 the stable release while 3.2XX are beta versions?

bill147 · Joined: 26 Oct 2004 Posts: 13

I have code that uses i2c and it does not work right with newer versions.

I have now found out why. The fast i2c speed is now faster than it was with compilers before 3.226. For instance the i2c_start() code with older compilers had 3 NOP's now there are no NOP's. If I use the Fast=200000 command my code now works with the new compiler versions.

Bill

Guest · Posted: Fri Jul 15, 2005 11:42 am

Sometimes it sound like they accidentally break something just to fix another. Trying not to do this is the hard part of programming sometimes.

I don't know if it would be realistic or not, but I would think they would have prepackaged test programs and boards to test each release with. Given the right design and tools, you could test every feature with every chip very quickly. This could help catch bugs without burdening customers too much by using them as debuggers. May also help a programmer catch a mistake when he/she made it instead of having to backtrack weeks or months later to find out how it got broken. Probably well worth it with the apparrent high frequency of releases along with the amount of chips supported.

Maybe they already do this, or something like it or better, and somthing else is the problem.

Mark · Joined: 07 Sep 2003 Posts: 2838 Location: Atlanta, GA

PCM Programmer has talked about this before:

http://www.webopedia.com/TERM/R/regression_testing.html