[Mono-devel-list] some mcs errors fixes

John Luke john.luke at gmail.com
Fri Oct 1 14:52:36 EDT 2004


Hello,

On Fri, 01 Oct 2004 14:12:11 +0530, Raja R Harinath
<rharinath at novell.com> wrote:
> Hi,
> 
> Marek Safar <marek.safar at seznam.cz> writes:
> 
> >>John Luke <john.luke at gmail.com> writes:
> 
> >>>Some of the error tests in mcs/errors are reporting the wrong error
> >>>because they either need to use -t:library or have a Main method.  The
> >>>attached patch adds one or the other depending on the test (sometimes
> >>>the compiler option would fail).  It makes one test pass and adds 19
> >>>tests to mcs-expect-no-error.  Is this ok to commit?
> >>
> > I hope this is little misunderstanding.
> 
> Not that I know of :-)
> 
> > Compilation of every single file
> > in this directory have to cause an error.
> > So, I am not sure whether this patch is useful.
> > Problem is that mcs compiler has not implemented all errors report and
> > this is the reason why exist *-expect-* files..
> 
> Yes.  However I believe there's a distinction between 'wrong-error' and
> 'no-error'.  'wrong-error' is less serious because we _do_ complain: we
> just don't complain the same way as CSC.  'no-error' is more serious
> since we accepts wrong code.
> 
> The cases John identified were spuriously reporting wrong-error just
> because we lacked a Main method.  These cases are more serious than they
> appeared to be.
> 
> - Hari
> 

Yes, Raja is correct and explained it better than I did originally.  I
fixed all the *-expect-errors files that he mentioned and will be
commiting in after make compiler-tests finishes, (I had been using
just make run-test in mcs/errors).  I have to say it is a little
tedious to have to modify 4 different files for each of these that
were incorrect, but perhaps it is not worth worrying about too much. 
Also, should we sort them by number or something, so they aren't
accidentily duplicated and easier to find.

I also have noticed some of the cs8*.cs tests seem to be either yield
tests that look incorrect and have equivalent working tests, or
generics tests that pass with gmcs, can someone in the know look at
them?



More information about the Mono-devel-list mailing list