[PATCH v5 2/2] io: Use gnulib fts implementation (BZ 22944, BZ 20331)

[Florian Weimer]

* Adhemerval Zanella Netto:

> On 13/04/26 12:53, Paul Eggert wrote:
>> On 2026-04-11 10:06, Florian Weimer wrote:
>>> I think we need to version the symbol, with an internal flag that
>>> does more stat'ing with FTS_NOSTAT.
>> 
>> Oh, let's not. FTS_NOSTAT means not to stat, and programs that rely
>> on FTS_NOSTAT to stat have always been broken. In the long run it's
>> obviously better to fix dracut etc. than to version the symbol, and
>> it seems to me that this is true even in the not-so-long run.
>
> Should we close BZ#34078 as WONTFIX then?
>
> [1] https://sourceware.org/bugzilla/show_bug.cgi?id=34078

I'm worried that more than just dracut is impacted.  The old behavior
had been around for a long time.

Maybe we can add compatibility symbols that perform post-processing in
fts_read and fts_children, calling the new implementation?  This way,
the necessary changes would be confined to glibc, without complicating
maintenance on the gnulib side.

Thanks,
Florian