On 11.10.2019 14:32, Greg Troxel wrote: > Kamil Rytarowski <n54%gmx.com@localhost> writes: > >> On 11.10.2019 05:42, Frédéric Fauberteau wrote: >>> Hi folks, >>> >> >> Is there any reason to use it on NetBSD? It silently breaks programs >> that pick iconv.h, as it overwrites /usr/include/iconv.h and later >> defines libiconv specific symbols that are absent in libc. >> >> This breaks e.g. GDB for ./build.sh tools >> >> I had bad experience with it and in my opinion it should be marked as >> not applicable/broken/unwanted for NetBSD. > > I don't follow this. In pkgsrc, there are often things that replicate > what's in NetBSD base. We have builtin.mk to decide whether to use > base or pkgsrc for things in pkgsrc that express dependencies on that > package. > > build.sh on NetBSD arguably shouldn't be looking in pkgsrc, and if it > does, it should use include/ and lib/ both. > > Can you explain "overwrites /usr/include/iconv.h"? I am not aware of > anything in pkgsrc writing to /usr/include, and if there is something, > that sounds like a bug to be fixed. > ./build.sh picks /usr/pkg/include (or $PREFIX) paths by default as they are detected by autoconf. I don't remember ofhand the reason for it, it could be pkg-config that is used by ./build.sh tools. It would be ideal to blacklist packages from pkgsrc that interfere with the process of bootstrapping tools. Before that, libiconv causes harm as /usr/pkg/include is in the search path, and prior /usr/include for headers with overlapping names. And in general, is there any reason to want libiconv on NetBSD? Why not to use the libc version one? If there is a missing feature, better to add it in libc, if it is non trivial and libiconv is a stop-gap for something in pkgsrc (I suffer because R picks it.. and stopped installing it recently), it would be good to first fix the the ./build.sh tools scripts.
Attachment:
signature.asc
Description: OpenPGP digital signature