On 23 Mar 2023, Steven Rostedt outgrape: > On Thu, 23 Mar 2023 12:43:29 -0700 > Luis Chamberlain <mcgrof@xxxxxxxxxx> wrote: > >> On Thu, Mar 23, 2023 at 04:45:02PM +0000, Nick Alcock wrote: >> > [Cc:ed Steve because I think he wanted something like this] >> > >> > % echo 1 > /sys/kernel/tracing/options/sym-unambiguous >> >> I'd like to really hear if Steven is really super excited about this or not. >> You keep saying he wants it, but I haven't him say it yet on the lists. > > I'm not "super excited" but I'm also not against it. Yeah, honestly I don't expect anyone to be *super* excited by any of this. It's not that big. It's not like, say, ftrace as a whole. It's just something that can be used to implement small improvements here and there. > But I don't like the > option name. Oh the name is dreadful! Better name suggestions much appreciated. > Perhaps "sym-file" ? ... Yes, except that it also decorates with built-in module names. > Or does this happen only when there's more than one function? Either way, > let's try to come up with something other than "sym-unambiguous". It happens only when at least one symbol in a given (object file * built-in module) pair would be ambiguous with respect to some other symbol with the same name without being given at least one of those two. It's a bit hard to pack into a couple of words... sym-unique is even worse, sym-objname is deeply unclear (what's an objname?) sym-filenames maybe? The question really is what property users care about, and I was hoping it would be that with this option turned on you can always tell all symbols apart from all other symbols. -- NULL && (void)
