Zach Welch wrote: > On Wed, 2009-07-08 at 12:47 +0200, Øyvind Harboe wrote: > >> How is the attached patch? >> > > The NEWS file details the migration of the installed scripts from > $(pkglibdir) to $(pkgdatadir)/scripts; the former should not be used > anymore, and custom scripts migrated into $(pkgdatadir)/site/. This > patch will prevent us from forcing this migration, as the old location > will continue to be be used. So it might fix the bug, but it seems bad. > > However, this particular example shows how our searching for files may > be insufficient. Presently, there is no way to enforce a separation > between script files and executable files (if such is desired). > The effect of this would be to allow a site to install read-only binary > files that may be used by read-write scripts. In reality, I think ACLs > make this proposition more meaningful, but others can comment on the > concept of privilege separation. Am I taking that idea too far here? > I do not see *that* as a problem, but I think binaries like the debug handler (or other modules/plugins/...) that are a fixed component of the installed OpenOCD version should only be loaded from $(pkglibdir): there is no use in looking for them in the user's scripts directory, and if a file with the right name but wrong contents is found there, it will even disturb operation.
> In this light, I would rather see us add a mechanism to find the two > types of file, such that the search paths for each are separated. > Script files can be searched using one list of paths, and binary files > from another. Internally, it seems like we could switch search paths > based on the filename extension being loaded. > Agreed, although I think the code that is looking for a file should specify the type of file, instead of relying on the extension. cu Michael _______________________________________________ Openocd-development mailing list [email protected] https://lists.berlios.de/mailman/listinfo/openocd-development
