Eli Zaretskii wrote: > > > That'd be a terrible > > > cycle killer, because this function would be called on each command > > > invocation. > > > > But no. Compared to what it takes the OS to spawn (and run!) a process, > > one path search more or less wont make even the slightest difference. > > How long is your Path? On a typical Windows machine, it tends to be > quite long, what with all kinds of packages that add to it at will. > It's nothing as short as "/usr/bin:/usr/local/bin" on Unix.
How long is my Path? ~200 bytes, 12 folders. > No, sorry, I cannot accept a solution that searches PATH before each > command. Perhaps we should introduce a new interface to the stored > variables, which would store a canonicalized (a.k.a. absolute) file > name (computed only once) for variables, like SHELL, that point to > file names. Then those parts of the code that need SHELL for running > programs could access that canonicalized file name instead of the > value of SHELL instead. Just to make shure I understand this: From a solution that is not even there you assume performance problems without any evidence, but have no idea either what to do with such problem in case it happens, nontheless suggest to transform this non-idea into a generic new interface for all variables "like SHELL"? Please explain. --- grischka _______________________________________________ Make-w32 mailing list Make-w32@gnu.org http://lists.gnu.org/mailman/listinfo/make-w32
