At 11:13 AM +0100 11/3/07, Rafael Garcia-Suarez wrote: >On 03/11/2007, John E. Malmberg <[EMAIL PROTECTED]> wrote: > > Some of the tests are trying to dynamically load modules on VMS.
In the one case I'm aware of (File::Find) it is the module itself rather than the test for it that uses dynamic loading, or, rather, uses a module -- Carp -- that itself uses dynamic loading. > > >> The TestInit.pm overrides @INC with a path that only works if the test >> does not change the default directory on VMS. >> >> This patch sets @INC to an absolute path so that changes of directories >> do not break dynamically loaded modules. > >I've applied a less invasive version of this patch, see below. >However, putting such a dependency on File::Spec is a bit disturbing, >even if limited only to VMS. Wouldn't it be better to fix the tests >instead ? (post-5.10 maybe) I don't particularly like it either, and I don't think it's a VMS-specific issue, but I'm not sure what else to suggest. The deadly combination is changing directories at run-time, having everything in @INC a relative path (such as when testing an uninstalled perl with t/TEST), and invoking a module that uses dynamic loading. When File::Find attempts to carp about something while running its tests, the only user-visible problem is: Can't locate Carp/Heavy.pm in @INC (@INC contains: ../lib ../lib) at ../lib/warnings.pm line 450. As far as I can tell, the only way in which this is a VMS problem is that we're the only ones triggering a warning. Perhaps for now we should have the File::Find tests put an absolute path in @INC and leave TestInit alone, though there are other modules, such as File::Path, that chdir all over the place and also use Carp. -- ________________________________________ Craig A. Berry mailto:[EMAIL PROTECTED] "... getting out of a sonnet is much more difficult than getting in." Brad Leithauser
