> > I'm not thrilled at the idea of going modify tens of C++ > > class header files, especially when this is a library > > frequently updated (POL). This is what I find insane. > > Multi-segmentation is perfectly handled by CW which does > > it properly, and I don't see any valid reason to have to > > declare where each function should go in the source code > > itself. This is like programming in the paleolithic era. > so, write a script? as john has mentioned before, you can do some > pre-processing to insert the __attribute__ tags :) the notion of > having a single section declaration for a single file has been > discussed, and, really.. it wont hurt to write a little scrip to > do the work (heck, it may already be done) :P
I agree with Florent, except that I far prefer the way Multilink does it to the way CodeWarrior does it. My optimal solution would be a cross between Multilink, GCC, and CodeWarrior. Just put functions into segments using a first-come-first-served algorithm, and if I want to force certain functions into certain segments, I'll tag the functions, or the file. The problem with your suggestion, Aaron, is that I don't know which segments the functions should go in, so I would have to take the time to figure out the segments, as well as write the script. Oh, yeah, and I've got 1908 C++ header files to change, not 10's, making the task that much less feasible. Finally, most of those 1908 files are shared between our PalmOS and our WinCE codebases, so the function attributes are just more noise that they'll have to wade through (and forget to include when they write new functions). The ability to tag each function, while useful, is at too low a level of granularity for me (and Florent) to use, so I've been using Multilink. (I've got a patch for the libraries it comes with, so if anyone is going to start maintaining it again, please email me for that.) I am very appreciative of the work that has gone in to the prc-tools, and since I haven't the skill to contribute, I'm want to make sure that people know that I'm not complaining, merely suggesting further directions the functionality might take to be more useful to me and the programmers I work with. (Also, if you have that script, or even a first cut, I'ld love to see it. How do you recognize a function definition that needs a tag? Would it work on in-lined constructors? Would it tag pure-virtual functions? I think it would be a hard function to write, but then, I'm just an applications guy... ;) Thanks, Blake. -- For information on using the Palm Developer Forums, or to unsubscribe, please see http://www.palmos.com/dev/support/forums/
