subcontext() Re: bitmaps as special enums?

Stefan (metze) Metzmacher metze at
Sun Feb 25 10:51:06 GMT 2007

Hash: SHA1

Jelmer Vernooij schrieb:
> On Sat, Feb 24, 2007 at 12:21:57PM +0100, Stefan (metze) Metzmacher wrote:
>> Jelmer Vernooij schrieb:
>>> On Fri, Feb 23, 2007 at 10:05:45AM +0100, Stefan (metze) Metzmacher wrote:
>>>> Jelmer Vernooij schrieb:
>>>>> On Fri, 2007-02-23 at 08:59 +0100, Stefan (metze) Metzmacher wrote:
>> please, don't change any existing subcontext() usages yet.
>> There're often very nasty details in the encoding and we some are
>> not tested in "make test".
>> Maybe try to start with the spoolss stuff only and make sure it still
>> works:
>> - - against windows
> If we have sufficient test coverage, why would this be necessary? The
> data that is sent by Windows shouldn't differ from the code that is
> being tested by our NDR testsuite. I do agree that testing against
> Windows in general is useful, but I think it'd be very hard for me to
> test all these functions...

All, want is that you fix represent_as for one specific case.

Then you should convert the rest step by step in sinple commits,
so that it's easier to review. And only cases which tests should be
converted. For other cases where we don't have tests, I'd like to have
tests first. (the drsuapi.idl and drsblobs.idl are the files a currently
care mostly about)

I think it would be usefull, when you let me convert some cases, with
your help via IRC. Then I can add some more tests:-)

BTW: the extra bytes in DsBind() seem to be bugs in windows to me.

you should find other examples here:

Also there are examples DsGetNCChanges() with MSZIP compression.

>> - - with ndrdump --dump --validate
> torture/ndr/ basically does what ndrdump --dump --validate does,
> but in an automated fashion. Why would running ndrdump be necessary if
> we have proper test coverage?

ok, it didn't know that that we do that same as --validate yet.
At least we should fail all spoolss tests where relative pointers are
used! As we need to have a max_seen_offset in the ndr_pull struct.
ndrdump currently give a warning of unread bytes. Also it compares
the original and generated buffer byte by byte.

If we do the same in the ndr tests than it's fine.

>> Also it would be very nice if we can avoid touching the callers of the
>> api.
> The callers won't have to be changed.


Version: GnuPG v1.4.5 (GNU/Linux)
Comment: Using GnuPG with SUSE -


More information about the samba-technical mailing list