Nathan Bush wrote: > Stephen Lau wrote: >> Nathan Bush wrote: >>> Stephen Lau wrote: >>>> Nathan Bush wrote: >>>>> Stephen Lau wrote: >>>>>> 2) We throw a "Header guard does not match filename" on a header >>>>>> guard that looks like: __FOO_H__ (for foo.h) or other things that >>>>>> don't strictly match _FOO_H_ >>>>>> Is this the best error message? Intuitively, I would think it >>>>>> should be "Invalid or missing header guard" myself - but this is >>>>>> the same behaviour as the gate's hdrchk.pl, so I'm curious what >>>>>> other people think. >>>>> >>>>> I think the two messages are used differently. "Invalid or missing >>>>> header >>>>> guard" is already used earlier in the code to indicate that the >>>>> "#ident" line >>>>> marking the start of the guard wasn't found where it was expected. >>>>> "Header >>>>> guard does not match filename" is used to indicate that the >>>>> constant value >>>>> found in the header does not match what was expected based on hdrchk's >>>>> application of cstyle rules. >>>> >>>> Yeah, I realise the different contexts of use. It just seems to me >>>> that __FOO_H__ not matching _FOO_H_ is a 'formatting' issue, rather >>>> than a 'filename' issue. It's minor, and not a huge deal to me though. >>> >>> I see your point. Perhaps you could reword the message slightly, >>> such as >>> changing "does not match filename" to "name does not follow suggested >>> style" >>> or something similar. >> >> What about: >> "Header guard does not match suggested style (_FILENAME_H_)" >> >> Or is that too nitty? > > No, not too nitty; it's helpful to state what is expected. But in your > example, is "_FILENAME_H_" a static string, or would you actually print > the real expected string based on the real filename? If so, in the case > of files like <sys/zone.h> you would want to suggest "_SYS_ZONE_H_" yet > you won't know about the "SYS_" part without looking at more than the > header basename. :(
I was going to just have it be a static string for that exact reason. :) maybe _FILEPATH_H_ would be better (more suggestive that the path should be incorporated) > I think that as long as the user gets a message that directs them to > review the style guidelines that should be good enough. Yeah. cheers, steve -- stephen lau // stevel at sun.com | 650.786.0845 | http://whacked.net opensolaris // solaris kernel development
