Thread Previous • Date Previous • Date Next • Thread Next |
On 14/04/16 08:23, Chris Johns wrote:
On 14/04/2016 16:09, Sebastian Huber wrote:On 14/04/16 08:04, Chris Johns wrote:On 14/04/2016 15:56, Sebastian Huber wrote:In case this is important, then we should add a test case to the RTEMS test suite.A test is a good idea. How would you write a test which checks all the possible externals used in newlib and I suppose gcc, indirectly via the apis or explicitly? I am thinking explicitly but how do we collect all the symbols and check they have no moved?It has to work automatically.I keep pressing the button but nothing is happening ... ok I will do it. :)I think a lot of symbols are missing in this file.I suppose the list is a unique set of all external symbols which do not internally resolve with in the tool set libraries. I suppose I can assume multlib variants all have the same unresolved symbols and can filter duplicates.I still not see how to write a suitable test. All you end up doing is testing the linker works and RTEMS is not missing any symbols. Any idea?
GCC uses link-time tests to check for operating system features. At least for libstdc++ some features are unsupported since some symbols are not defined (only declared in header files) during GCC build. Let U be the set of all undefined symbols in libc, libm and libgcc. Let D be the set of all defined symbols in libc, libm, libgcc. Then U \ D must go into crt0.c. Also, these symbols must be defined by librtemscpu or librtemsbsp.
-- Sebastian Huber, embedded brains GmbH Address : Dornierstr. 4, D-82178 Puchheim, Germany Phone : +49 89 189 47 41-16 Fax : +49 89 189 47 41-09 E-Mail : sebastian.huber@xxxxxxxxxxxxxxxxxx PGP : Public key available on request. Diese Nachricht ist keine geschäftliche Mitteilung im Sinne des EHUG. _______________________________________________ users mailing list users@xxxxxxxxx http://lists.rtems.org/mailman/listinfo/users
Thread Previous • Date Previous • Date Next • Thread Next |