Amos Jeffries wrote:
Alex Rousskov wrote:
On Sat, 2008-02-02 at 13:12 +0000, Amos Jeffries wrote:
Update of cvs.devel.squid-cache.org:/cvsroot/squid/squid3/src
Modified Files:
Tag: ayjwork
HttpRequestMethod.cc HttpRequestMethod.h SquidString.h
String.cci Log Message:
Compile errors. From HttpRequestMethod class addition.
Test failures are occuring and still need fixing.
Amos,
I could not tell from the diff what went wrong (too much formatting
noise?).
Several compiling problems at once when "make check" is run. Try it on
HEAD to see what I mean if you like:
- operator << overloading problem (more on that below).
- CPPUNIT_ASSERT_EQUALS requires two parameters of identical types. It
will NOT do any silent conversion. So _method_t vs HTTPRequestMethod
tests all failed building.
- Const_str() method changed to image() but some calls left unaltered.
Please let me know if you need any help with HttpRequestMethod
polishing other than the test cases. If am especially worried about
this:
+ #if 0
+ /* AYJ: Duplicated code. Seem an HttpRequestMethod can itself be
converted to String() and
+ * this operator clashes with that operator.
+ */
inline std::ostream &
operator << (std::ostream &os, HttpRequestMethod const &method)
***************
*** 171,174 ****
--- 173,177 ----
return os;
}
+ #endif
I do not see how HttpRequestMethod can be converted into a String, but
if that is the case, we should probably remove such auto-conversion.
I do not know what error you are getting, but it is possible that
auto-conversion from HttpRequestMethod to method_t is at fault. If so,
we should probably try to remove that conversion.
Yes it was, I found that in the tests hat were still failing it keeps
displaying HttpRequestMethod("GET") as '1' instead of "GET".
For some weird reason my compiler was saying a bad overloading between:
operator <<(ostream &os, String const &s)
and operator <<(ostream &os, HttpRequestMethod const &s)
which threw me for a while.
Dropping the silent _method_t / int conversion proves to be a better
solution than dropping the <<. I'm testing making access to the int
value an explicit call to value() instead of silent conversion.
Again, let me know if you need a hand.
After my next patch goes into ayjwork branch, if you wouldn't mind
running every test you know on the branch it would be good. :-)
But as usual I'm not expecting anyone but me will go near the code until
its already in HEAD. :-(
Sigh, it looked promising fro a while.
The exact errors is when compiling testHeaders_HttpRequestMethod.cc.
The test where there is no code run, just the dependency and inline build.
./HttpRequestMethod.h: In function ‘std::ostream&
operator<<(std::ostream&, const HttpRequestMethod&)’:
./HttpRequestMethod.h:181: error: ambiguous overload for ‘operator<<’ in
‘os << ((const HttpRequestMethod*)method)->HttpRequestMethod::image()’
./String.cci:186: note: candidates are: std::ostream&
operator<<(std::ostream&, const String&)
./HttpRequestMethod.h:179: note: std::ostream&
operator<<(std::ostream&, const HttpRequestMethod&)
And its back, removing the silent _method_t() conversion has no effect.
I'll keep experimenting, but if you have any ideas please let me know.
Amos
--
Please use Squid 2.6STABLE17+ or 3.0STABLE1+
There are serious security advisories out on all earlier releases.
