Re: Discussion about why GNU/Linux system upgrades cause old programs to break

> From: Arsen Arsenović <[email protected]>
> Cc: [email protected],  [email protected]
> Date: Fri, 14 Aug 2026 00:12:29 +0200
> 
> Eli Zaretskii <[email protected]> writes:
> 
> > Please consider the plight of the person who bumps into these problems
> > and needs to solve them so the program (which built and worked for
> > ages) will build again.  At the very least, they need to:
> >
> >   . realize such an option exists
> >     - indexing of the manual is not a great help here: there are no
> >       index entries in the Concept Index leading directly to it; a
> >       few entries that mention some of the relevant warnings do lead
> >       to the "Warning Options" node, but the node is very long
> >   . read and understand the description of each of the -Wfoo warnings
> >     it disables and decide whether it's applicable and if so, whether
> >     it's desirable to disable it
> >   . preferably understand the significance of C Standard versions to
> >     which the manual alludes, and their relation to the program's code
> >
> > This requires non-trivial knowledge of the C Standard and related
> > subjects, and also non-trivial amount of tenacity, let alone time and
> > effort.
> >
> > If this is not evidence how poorly we treat users in such situations,
> > I don't know what is.
> 
> As I've said elsewhere, I think that this particular change has helped
> many orders of magnitude more users than it has harmed.

You have been presented with several practical cases where this
decision causes trouble to people who use GCC correctly.  And you
still keep claiming that being able to "help" many more clueless
people who couldn't have been bothered to RTFM to discover -pedantic
somehow trumps that.  In what world can this be a valid stance?

> I do not object to the idea that -fpermissive might need to get featured
> more prominently in the documentation.  That may well be true.
> 
> It may even be useful to emit a hint about it as I've stated elsewhere.

I submit that the fact GCC currently doesn't is because GCC developers
overwhelmingly think this is a non-issue.

> But, as I've also stated elsewhere, I highly doubt that knowing about
> -fpermissive would've actually saved many here.

Why is that doubt even an issue in this context?  If there's something
that can be done to enhance the documentation to help people in these
situation, it should be done, full stop.