From: Mike H. <mi...@th...> - 2004-01-03 20:45:43
|
Hi. Yes, I saw your patch (but only after I had written mine). > My patch didn't do this, I think Ethan said that this is not necessary. I did this at the suggestion of Christian, who sometimes runs gaim in tree apparently (or he said other devs do). > How is this approach better than Ben's or mine? (I've not looked at his patch, > I assume that it's > quite functionally similar, if not the same, as mine) His works the same way, basically. I don't mean to diss your work, it's good, but I think my patch works better on Linux because: - It doesn't work by scanning the path. It actually gets the absolute path of the binary from the kernel. That makes it Linux specific, but more accurate/reliable. Bens patch is merged too so Solaris/other UNIX users who want this can get a similar type of functionality. - It works by redefining DATADIR, LIBDIR etc so the patch is a bit simpler to review. We want to avoid attempting to figure out where the binary is by the technique used in your patch and Bens because it's a bit inefficient, it doesn't work if the binary isn't in the path, and it can get it wrong - for instance I think if . is in the PATH .... not sure though, it might be as accurate as the mmaps approach. > I am attaching this outdated patch for reference purpose. It was tested on OS > X and Linux. One question - I don't understand how this line compiles: - LIBDIR "/libnssckbi.so", + gaim_get_lib_dir() "/libnssckbi.so", You seem to be using a preprocessor string join, but gaim_get_lib_dir() is a function?? How does that work? thanks -mike |