> (I believe an empty GNUSTEP_CONFIG_FILE should cause no config file to > be read and the compiled-in defaults to be used with no warnings, but I > suppose we need to check if that really works).
Actually, that's not how it works currently, at least not gnustep-make. If GNUSTEP_CONFIG_FILE is not set at all, we want/need the default config file to be used ... if GNUSTEP_CONFIG_FILE is set to be empty, well, we could then not read any, but I'm not sure I remember a way in sh to make a difference between variables that are not set vs. ones that are set to an empty value ... is there one (and is it portable) ? :-P It would have been nice to make GNUSTEP_CONFIG_FILE and GNUSTEP_USER_CONFIG_FILE consistent in this respect, but it looks hard - actually I'm no longer sure GNUSTEP_USER_CONFIG_FILE works/can work properly with the current empty value semantics (currently, you disable user config files by doing GNUSTEP_USER_CONFIG_FILE=<nothing>). Eg, if the system config file is not found, the default GNUSTEP_USER_CONFIG_FILE should be used ... it's not at the moment, and to fix it we're back at making a difference in sh between a variable that is not set versus one that's empty. :-( Maybe to disable config files you'd set the corresponding variable / config value to a non-existing or empty file. I suppose that might work, it's easy to ignore non-existing or empty files. ;-) Suggestions welcome, thanks :-) _______________________________________________ Bug-gnustep mailing list [email protected] http://lists.gnu.org/mailman/listinfo/bug-gnustep
