On 2017-05-06 05:03, Jon Turney wrote: > On 05/05/2017 21:42, Yaakov Selkowitz wrote: >> On 2017-05-02 08:13, Jon Turney wrote: >>> Converting a dependency atom to a package name with full generality >>> requires >>> a database of all the pathnames contained in all packages, so we don't >>> even >>> try to do that here, leaving that task for downstream processing of the >>> .hint file... >> Or should we just replace DEPEND, which allowed atoms which are >> unsupported by setup and whose name is confusing wrt REQUIRES, with a >> BUILDREQUIRES variable which is package-name only? > Changing the name to something less confusing seems like a good > idea. > I kind of like the current scheme, as it can do some checks when > cross-compiling as well. > I think that something (calm?) needs to keep a database of all > package pathnames, to check for pathname collisions between packages > (I suspect we have some unnoticed cases of that at the moment), so > the extra cost of doing a conversion from cygport dependency atoms in > a .hint to package names in setup.ini shouldn't be too much. It'll be > a while before that happens, though... :) > People also like to typo this as DEPENDS, so some sort of warning > that the .cygport sets a variable which has no effect might also be > useful.
Append to DEPEND with a warning? -- Take care. Thanks, Brian Inglis, Calgary, Alberta, Canada