On Fri, Mar 6, 2020 at 9:16 AM Daniel Burgener <dburgener@xxxxxxxxxxxxxxxxxxx> wrote: > > On 3/6/20 8:57 AM, Stephen Smalley wrote: > > On Thu, Mar 5, 2020 at 1:40 PM Daniel Burgener > > <dburgener@xxxxxxxxxxxxxxxxxxx> wrote: > >> When the lexer encounters an unexpected character in a policy source > >> file, it prints a warning, discards the character and moves on. In > >> some build environments, these characters could be a symptom of an > >> earlier problem, such as unintended results of expansion of > >> preprocessor macros, and the ability to have the compiler halt on > >> such issues would be helpful for diagnosis. Signed-off-by: Daniel > >> Burgener <Daniel.Burgener@xxxxxxxxxxxxx> > > I'm trying to remember why this particular case (unmatched character > > in the lexer) isn't already a fatal error. If there isn't a real > > reason for it, we could alternatively just switch it to use yyerror() > > in that case. Otherwise, your description suggests that you only want > > to make that particular case a fatal error; are you sure you want to > > treat all warnings as fatal errors? > > That's a bug in the description rather than the code. That particular > case is what caused me to want this option, but my intent was to add an > ability to treat all warnings as errors. I can resubmit with a better > message if you'd like. > > I'm not sure why the behavior is that that particular case is a > warning. Git blame shows it as going back to the import from svn, so > it's been that way for a while. I'd think that --werror would be > helpful either way. If no one has a reason for this particular case to > be a warning rather than an error, I can submit a patch for that change > as well. Ok. I have no objections to your patch, just wanted to make sure it is doing what you really want it to do. I'd also be willing to accept a patch to make unexpected characters fatal.
