DTrace: document the "nolibs" option

[domagoj.stolfa_gmail.com]

Summary

This diff just extends the dtrace(1) man page to mention -x nolibs.

[allanjude]

cddl/contrib/opensolaris/cmd/dtrace/dtrace.1

608

It might be worth mentioning the /usr/lib/dtrace path, and that this can be used to resolve the error jhb was running into.

Infact, it might be worth putting the exact error message here, so that search results for that error lead to this man page.

[markj]

Thanks. Would you also consider documenting -x libdir and -x syslibdir while here?

[domagoj.stolfa_gmail.com]

Added -x libdir and -x syslibdir as suggested by @markj

[domagoj.stolfa_gmail.com]

cddl/contrib/opensolaris/cmd/dtrace/dtrace.1

608

It might be worth doing -- will update the diff again once jhb gives me the exact error message.

[markj]

cddl/contrib/opensolaris/cmd/dtrace/dtrace.1

635

"Defaults to .Pa /usr/lib/dtrace ."

[markj]

cddl/contrib/opensolaris/cmd/dtrace/dtrace.1

608

There are lots of error messages that can arise when parsing scripts in /usr/lib/dtrace.

[domagoj.stolfa_gmail.com]

@markj 's comments.

cddl/contrib/opensolaris/cmd/dtrace/dtrace.1

608

That's true -- and the error message in this particular case doesn't really seem like a great one to begin with:

dtrace: invalid probe specifier fbt::vn_sendfile:return { trace(arg1) }: "/usr/lib/dtrace/io.d", line 1: cannot find type: struct devstat*: Type information is in parent and unavailable

Perhaps if we want to avoid confusion here, the right thing is to change the "invalid probe specifier" error message to something more meaningful?

[markj]

cddl/contrib/opensolaris/cmd/dtrace/dtrace.1

608

This is definitely something we can do better... I'll just say that the error handling in libdtrace is really quite hard to follow, there's some setjmp/longjmp and complicated interactions with the generated lexer and parser. So it may be an easy change, but also quite possibly is not an easy change.

637