[libcamera-devel] [PATCH v2] meson: Don't set _FORTIFY_SOURCE for ChromeOS

[George Burgess]

Hey,

> Do you expect to have -D_FORTIFY_SOURCE and -U_FORTIFY_SOURCE in the
> same command line, so that you have to reset has_fortify_define to
> false if it was previously been set to true here below ?

You're correct - the code as written intends to answer "will FORTIFY _be
enabled_ by `cpp_args`?" If you'd like, I'm happy to reduce complexity and have
the code determine whether `-D_FORTIFY_SOURCE*` was specified at all. Either
would work equally well for ChromeOS. :)

George

On Tue, Sep 12, 2023 at 7:22 AM Jacopo Mondi
<jacopo.mondi at ideasonboard.com> wrote:
>
> Hi George
>
> On Mon, Sep 11, 2023 at 05:09:07PM -0600, George Burgess IV via libcamera-devel wrote:
> > ChromeOS is moving to a platform default of `_FORTIFY_SOURCE=3`, and
> > this definition conflicts with that:
> >
> > <command line>:4:9: error: '_FORTIFY_SOURCE' macro redefined
> > [-Werror,-Wmacro-redefined]
> >
> > Rather than adding logic to keep up with their local configuration, it
> > seems best to leave setting _FORTIFY_SOURCE on ChromeOS up to ChromeOS.
> >
> > Signed-off-by: George Burgess IV <gbiv at google.com>
> > ---
> >  meson.build | 24 +++++++++++++++++++-----
> >  1 file changed, 19 insertions(+), 5 deletions(-)
> >
> > diff --git a/meson.build b/meson.build
> > index 7959b538..2e834263 100644
> > --- a/meson.build
> > +++ b/meson.build
> > @@ -99,12 +99,26 @@ if cc.get_id() == 'clang'
> >          error('clang version is too old, libcamera requires 9.0 or newer')
> >      endif
> >
> > -    # Turn _FORTIFY_SOURCE by default on optimised builds (as it requires -O1
> > -    # or higher). This is needed on clang only as gcc enables it by default.
> > +    # Turn _FORTIFY_SOURCE by default on. This is needed on clang only as gcc
> > +    # enables it by default. FORTIFY will not work properly with `-O0`, and may
> > +    # result in macro redefinition errors if the user already has a setting for
> > +    # `-D_FORTIFY_SOURCE`. Do not enable FORTIFY in either of those cases.
> >      if get_option('optimization') != '0'
> > -        common_arguments += [
> > -            '-D_FORTIFY_SOURCE=2',
> > -        ]
> > +        has_fortify_define = false
> > +        # Assume that if the user requests a FORTIFY level in cpp_args, they
> > +        # do the same for c_args.
> > +        foreach flag : get_option('cpp_args')
> > +            if flag == '-U_FORTIFY_SOURCE'
> > +                has_fortify_define = false
>
> Do you expect to have -D_FORTIFY_SOURCE and -U_FORTIFY_SOURCE in the
> same command line, so that you have to reset has_fortify_define to
> false if it was previously been set to true here below ?
>
> Otherwise I might be missing why you have to set 'has_fortify_define =
> false' when the variable is already initialized to 'false'...
>
> > +            elif flag.startswith('-D_FORTIFY_SOURCE=')
> > +                has_fortify_define = true
> > +            endif
> > +        endforeach
> > +        if not has_fortify_define
> > +            common_arguments += [
> > +                '-D_FORTIFY_SOURCE=2',
> > +            ]
> > +        endif
> >      endif
> >
> >      # Use libc++ by default if available instead of libstdc++ when compiling
> > --
> > 2.42.0.283.g2d96d420d3-goog
> >