Re: [PATCH] bpf: be less picky about the content of bpf_helper_defs.h

[Kris Van Hees <kris.van.hees@oracle.com>]

On Mon, May 06, 2024 at 03:43:19PM -0400, Eugene Loh wrote:
> Reviewed-by: Eugene Loh <eugene.loh@oracle.com>
> 
> One possibility, however, is to regularize the input line with regards to
> spaces and tabs.  E.g., start with
> 
> { gsub("[ \t]+", " ") }
> 
> Then, subsequent string processing can forget about that \t stuff and will
> presumably be that much more simple and readable.

Yes, but means we modify every line, which we do not really need to do.  I
guess it's one trade-off vs another.  The gsub you mention only really helps
drop \t becauxe you still need to account for places where whitespace is
optional.

> On 5/6/24 15:05, Kris Van Hees wrote:
> > There are small differences in the format of the BPF helper definitions
> > in bpf_helper_defs.h in libbpf headers.  This patch allows mkHelpers to
> > be more accepting of these differences.
> > 
> > Signed-off-by: Kris Van Hees <kris.van.hees@oracle.com>
> > ---
> >   include/mkHelpers | 15 ++++++++-------
> >   1 file changed, 8 insertions(+), 7 deletions(-)
> > 
> > diff --git a/include/mkHelpers b/include/mkHelpers
> > index 1a7537ca..dc5e62c8 100755
> > --- a/include/mkHelpers
> > +++ b/include/mkHelpers
> > @@ -8,14 +8,15 @@ BEGIN {
> >   }
> >   $1 == "static" && /=/ && /[ \t]+[1-9][0-9]*[ \t]*;/ {
> > -	match($0, /[ \t]+[1-9][0-9]*[ \t]*;/);
> > -	id = substr($0, RSTART, RLENGTH - 1);
> > -	gsub(/[ \t]+/,"", id);
> > -	match($0, /\([ \t]*\*[ \t]*bpf_[_A-Za-z0-9]+/);
> > -	fn = substr($0, RSTART + 1, RLENGTH - 1);
> > -	sub(/[ \t]*\*[ \t]*bpf_/, "", fn);
> > +	if (match($0, /[ \t\*]bpf_[_A-Za-z0-9]+/) > 0) {
> > +		fn = substr($0, RSTART + 5, RLENGTH - 5);
> > -	print "#define BPF_FUNC_"fn " " id;
> > +		match($0, /[ \t]+[1-9][0-9]*[ \t]*;/);
> > +		id = substr($0, RSTART, RLENGTH - 1);
> > +		gsub(/[ \t]+/,"", id);
> > +
> > +		print "#define BPF_FUNC_"fn " " id;
> > +	}
> >   }
> >   END {