* Re: [musl] fts.h
2025-01-27 15:25 ` [musl] fts.h Rich Felker
@ 2025-01-27 15:34 ` enh
2025-01-27 19:17 ` Khem Raj
1 sibling, 0 replies; 5+ messages in thread
From: enh @ 2025-01-27 15:34 UTC (permalink / raw)
To: Rich Felker; +Cc: musl, Aditya Kumar
On Mon, Jan 27, 2025 at 10:25 AM Rich Felker <dalias@libc.org> wrote:
>
> On Thu, Jan 23, 2025 at 01:54:41PM -0500, enh wrote:
> > https://wiki.musl-libc.org/faq says "If glibc bug 15838 is fixed by
> > adding an fts64 interface in glibc, we could consider supporting it
> > with a matching ABI in musl, but it seems more likely that glibc will
> > just deprecate this interface", but that bug _was_ fixed in 2015 for
> > glibc 2.23...
>
> I wonder when that text was written. While we could certainly consider
> it, lack of any apparent need so far suggests that it wouldn't meet
> the modern criteria for inclusion in musl.
well, the reason i'm mentioning this is because Android's mixing in
the bionic (lightly modified openbsd) fts to its host musl to be able
to build various things.
one of them is the bionic fts test (we build the bionic tests against
the host libc too, for comparison), so we can ignore that.
bionic/tests/fts_test.cpp:32:10: fatal error: 'fts.h' file not found
the other two seem legit though:
external/vboot_reference/futility/updater_archive.c:13:10: fatal
error: 'fts.h' file not found
external/selinux/libselinux/utils/selabel_get_digests_all_partial_matches.c:7:10:
fatal error: 'fts.h' file not found
> The main motivation I could potentially see flipping this is if there
> are a significant number of programs shipping their own (e.g. gnulib?)
> versions of fts, that would save significant code-duplication disk
> space (or get better behavior of some sort) if using a shared copy in
> libc.
iirc there's a way to grep the source of all debian packages, though
it would be hard to know how many of them have an alternative like
ftw(3) they can use instead.
> Rich
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: [musl] fts.h
2025-01-27 15:25 ` [musl] fts.h Rich Felker
2025-01-27 15:34 ` enh
@ 2025-01-27 19:17 ` Khem Raj
1 sibling, 0 replies; 5+ messages in thread
From: Khem Raj @ 2025-01-27 19:17 UTC (permalink / raw)
To: musl; +Cc: enh, Aditya Kumar
On Mon, Jan 27, 2025 at 7:25 AM Rich Felker <dalias@libc.org> wrote:
>
> On Thu, Jan 23, 2025 at 01:54:41PM -0500, enh wrote:
> > https://wiki.musl-libc.org/faq says "If glibc bug 15838 is fixed by
> > adding an fts64 interface in glibc, we could consider supporting it
> > with a matching ABI in musl, but it seems more likely that glibc will
> > just deprecate this interface", but that bug _was_ fixed in 2015 for
> > glibc 2.23...
>
> I wonder when that text was written. While we could certainly consider
> it, lack of any apparent need so far suggests that it wouldn't meet
> the modern criteria for inclusion in musl.
>
> The main motivation I could potentially see flipping this is if there
> are a significant number of programs shipping their own (e.g. gnulib?)
> versions of fts, that would save significant code-duplication disk
> space (or get better behavior of some sort) if using a shared copy in
> libc.
In yocto, we use the fts library https://github.com/pullmoll/musl-fts
and checked core layer and meta-openembedded layer which is 3000+ packages
following 8 are depending on it explicitly.
pmdk, fluentbit, libabigail, dracut, overlayfs-tools, libcgroup, ltp elfutils
>
> Rich
^ permalink raw reply [flat|nested] 5+ messages in thread