On 13/05/18 12:53, Eli Schwartz wrote: > On 05/12/2018 07:17 AM, Allan McRae wrote: >> On 03/05/18 11:32, Eli Schwartz wrote: >>> This is a common URI scheme (in general if not in makepg) and we should >>> provide a handler for it. We already allow its use for locally sourced >>> git repositories, so it makes sense to not leave files out. >>> >>> Signed-off-by: Eli Schwartz <[email protected]> --- >>> etc/makepkg.conf.in | 3 ++- >>> 1 file changed, 2 insertions(+), 1 deletion(-) >>> >>> diff --git a/etc/makepkg.conf.in b/etc/makepkg.conf.in >>> index fe3858ad..4d6ab78d 100644 >>> --- a/etc/makepkg.conf.in >>> +++ b/etc/makepkg.conf.in >>> @@ -8,7 +8,8 @@ >>> # >>> #-- The download utilities that makepkg should use to acquire sources >>> # Format: 'protocol::agent' >>> -DLAGENTS=('ftp::/usr/bin/curl -gqfC - --ftp-pasv --retry 3 --retry-delay 3 >>> -o %o %u' >>> +DLAGENTS=('file::/usr/bin/curl -gqC - -o %o %u' >> >> why not use "cp"? > > 1) no need to strip the file:/// bit by special-casing file:// like we > do scp. Why make file:// filesystem-aware in ways that nothing else is > anyway? > > 2) cp preserves attributes like the executable bit, which http urls > specifically don't; cp is therefore inconsistent. > > 3) cp is silent, which would make it the only DLAGENTS which is by > default silent. > > 4) it's thematically consistent with using the same thing lots of times. > > I guess we could use /usr/bin/cp --no-preserve=mode -v > > But this still leaves special-casing file:// which IMHO introduces > slightly magic behavior and I don't like using something other than the > actual contents of the source=() except as explicitly documented in the > PKGBUILD(5) manpage for :: and + >
Good enough explanation for me. A
