------- Comment #3 from kargl at gcc dot gnu dot org 2009-04-22 05:52 ------- (In reply to comment #2) > no quite. > but the std post on limitation on this , so I really hope gfortran support it. > > http://groups.google.com/group/comp.lang.fortran/browse_thread/thread/f4ab93c7cece56ee/d4518a395a0fd4fe?hl=zh-CN#d4518a395a0fd4fe > > today I check the Changelog ,I found > * resolve.c (resolve_global_procedure): Enable whole-file checking for > procedures that are declared later in the file. > > so I post this to gcc bugzilla. > > can the developers think about this? >
Given the URL you posted, I'm inclined to say this should be closed with WONTFIX. It is a clear violation of the standard: Note 2.2 in F2003 states: This standard places no ordering requirement on the program units that constitute a program, butmbecause the public portions of a module are required to be available by the time a module reference (11.2.1) is processed, a processor may require a particular order of processing of the program units. In 11.2.1, one finds The USE statement specifies use association. A USE statement is a module reference to the module it specifies. At the time a USE statement is processed, the public portions of the specified module shall be available. When the USE AFTER is processed, the public is not available. period. Is really that difficult to re-order the code? In particular, how many other compiler accept your code? -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=39844
