[1.35.0] Test regressions
For a full list of 1.35.0 test regressions, see http://beta.boost.org/development/tests/release/developer/summary.html There are a few 1.35.0 regressions I'm not particularly concerned about; VC++ 7.1 is now old news, and no one seems to care much about the Intel/Windows compiler. Given that VC++ 9.0 has been out for awhile now, it would be nice to kill off a few more of those regressions before 1.35.0 ships: * Conversion * Graph * Python * TR1 Red regressions I'm concerned about are: * Numeric/interval on Darwin. * Serialization on Darwin. * Spirit on Darwin. If anyone can quickly come up with fixes, there is still (barely) enough time to get them into 1.35.0. --Beman
On Mar 19, 2008, at 9:42 AM, Beman Dawes wrote:
For a full list of 1.35.0 test regressions, see http://beta.boost.org/development/tests/release/developer/summary.html
Or, better yet: http://beta.boost.org/development/tests/release/developer/issues.html
There are a few 1.35.0 regressions I'm not particularly concerned about; VC++ 7.1 is now old news, and no one seems to care much about the Intel/Windows compiler.
Given that VC++ 9.0 has been out for awhile now, it would be nice to kill off a few more of those regressions before 1.35.0 ships:
* Graph
As Richard Webb noted, it fails for the same reason as VC++ 8.0 fails, so we should just mark this as "expected" as well. Okay for me to go ahead and do that on the trunk and release branch?
Red regressions I'm concerned about are:
* Numeric/interval on Darwin. * Serialization on Darwin.
This is the trivial patch to fix the Serialization failures on Darwin: Index: libs/serialization/example/portable_binary_oarchive.hpp =================================================================== --- libs/serialization/example/portable_binary_oarchive.hpp (revision 43724) +++ libs/serialization/example/portable_binary_oarchive.hpp (working copy) @@ -52,7 +52,7 @@ public: friend class boost::archive::basic_binary_oarchive<portable_binary_oarchive>; friend class boost::archive::save_access; #endif - void save_impl(const long l){ + void save_impl(long l){ long ll = l; char size = 0; if(l < 0){ Tested on Darwin (both PowerPC and Intel); shall I commit to the trunk and the release branch? - Doug
Doug Gregor <dgregor <at> osl.iu.edu> writes:
There are a few 1.35.0 regressions I'm not particularly concerned about; VC++ 7.1 is now old news, and no one seems to care much about the Intel/Windows compiler.
Given that VC++ 9.0 has been out for awhile now, it would be nice to kill off a few more of those regressions before 1.35.0 ships:
* Graph
As Richard Webb noted, it fails for the same reason as VC++ 8.0 fails, so we should just mark this as "expected" as well. Okay for me to go ahead and do that on the trunk and release branch?
As far as the other VC9 failures go: * Conversion The lexical_cast_loopback_test test is marked as an expected failure on both VC7.1/8.0. I guess that VC9 has the same problem. * Python I think this is a problem in the tests rather than the lib itself. See: http://svn.boost.org/trac/boost/ticket/1516 * TR1 The test_hash\std_test_hash failures are a bug in VC9 - see my post @ http://article.gmane.org/gmane.comp.lib.boost.devel/172149 For VC7.1: The Fusion failures are a bit odd - the tests pass on Trunk, and pass on Release if i run the tests manually. I don't know what causes the failures. Thanks, Richard Webb
Doug Gregor wrote:
On Mar 19, 2008, at 9:42 AM, Beman Dawes wrote:
For a full list of 1.35.0 test regressions, see http://beta.boost.org/development/tests/release/developer/summary.html
Or, better yet:
http://beta.boost.org/development/tests/release/developer/issues.html
There are a few 1.35.0 regressions I'm not particularly concerned about; VC++ 7.1 is now old news, and no one seems to care much about the Intel/Windows compiler.
Given that VC++ 9.0 has been out for awhile now, it would be nice to kill off a few more of those regressions before 1.35.0 ships:
* Graph
As Richard Webb noted, it fails for the same reason as VC++ 8.0 fails, so we should just mark this as "expected" as well. Okay for me to go ahead and do that on the trunk and release branch?
Yes, please!
Red regressions I'm concerned about are:
* Numeric/interval on Darwin. * Serialization on Darwin.
This is the trivial patch to fix the Serialization failures on Darwin:
Index: libs/serialization/example/portable_binary_oarchive.hpp =================================================================== --- libs/serialization/example/portable_binary_oarchive.hpp (revision 43724) +++ libs/serialization/example/portable_binary_oarchive.hpp (working copy) @@ -52,7 +52,7 @@ public: friend class boost::archive::basic_binary_oarchive<portable_binary_oarchive>; friend class boost::archive::save_access; #endif - void save_impl(const long l){ + void save_impl(long l){ long ll = l; char size = 0; if(l < 0){
Tested on Darwin (both PowerPC and Intel); shall I commit to the trunk and the release branch?
Yes, thanks! --Beman
On Mar 19, 2008, at 11:58 AM, Beman Dawes wrote:
Doug Gregor wrote:
As Richard Webb noted, it fails for the same reason as VC++ 8.0 fails, so we should just mark this as "expected" as well. Okay for me to go ahead and do that on the trunk and release branch?
Yes, please!
Done, along with the Conversion library markup.
Tested on Darwin (both PowerPC and Intel); shall I commit to the trunk and the release branch?
Yes, thanks!
Committed to the release branch. There seem to be other issues on the trunk with this test. - Doug
Actually, this doesn't fix anything. Since changes to binary archives where checked in, any archive classes derived from this archive are no longer guarenteed to work. Portable archive is such a class. In fact, the original intent of the example was to illustrate how to make a new archive by deriving from an existing one. The changes to binary archive now render this example incorrect and misleading. I would not now recommend using binary archive as a base class for this reason. The new version of portable archive - recently checked into the trunk - fixes this by not being based on binary archive any more. It's been tested to create portable archives - but its still being checked out. Unfortunately, this was not discovered until late in the developement cycle and its not a trivial fix. It ripples through the documentation. And of course, we need a new example to illustrate how best to do this. And this entails making a good example, and test and corresponding documentation. So its not a trivial fix. The only fix at this point is to update the Introduction section to include the information related above. Robert Doug Gregor wrote:
On Mar 19, 2008, at 9:42 AM, Beman Dawes wrote:
For a full list of 1.35.0 test regressions, see http://beta.boost.org/development/tests/release/developer/summary.html
Or, better yet:
http://beta.boost.org/development/tests/release/developer/issues.html
There are a few 1.35.0 regressions I'm not particularly concerned about; VC++ 7.1 is now old news, and no one seems to care much about the Intel/Windows compiler.
Given that VC++ 9.0 has been out for awhile now, it would be nice to kill off a few more of those regressions before 1.35.0 ships:
* Graph
As Richard Webb noted, it fails for the same reason as VC++ 8.0 fails, so we should just mark this as "expected" as well. Okay for me to go ahead and do that on the trunk and release branch?
Red regressions I'm concerned about are:
* Numeric/interval on Darwin. * Serialization on Darwin.
This is the trivial patch to fix the Serialization failures on Darwin:
Index: libs/serialization/example/portable_binary_oarchive.hpp =================================================================== --- libs/serialization/example/portable_binary_oarchive.hpp (revision 43724) +++ libs/serialization/example/portable_binary_oarchive.hpp (working copy) @@ -52,7 +52,7 @@ public: friend class boost::archive::basic_binary_oarchive<portable_binary_oarchive>; friend class boost::archive::save_access; #endif - void save_impl(const long l){ + void save_impl(long l){ long ll = l; char size = 0; if(l < 0){
Tested on Darwin (both PowerPC and Intel); shall I commit to the trunk and the release branch?
- Doug _______________________________________________ Unsubscribe & other changes: http://lists.boost.org/mailman/listinfo.cgi/boost
On Mar 19, 2008, at 1:42 PM, Robert Ramey wrote:
Actually, this doesn't fix anything.
Well, it makes the test build and execute.
The only fix at this point is to update the Introduction section to include the information related above.
That's up to you and the release manager. - Doug
participants (4)
-
Beman Dawes -
Doug Gregor -
Richard Webb -
Robert Ramey