After a fresh install of beta2 on a laptop (trident cyberblade / lcd combo), starting X produces a lot of errors. pex5 and xie modules fail to load, due to undefined symbols (PexExtensionInitPtr and XieInitPtr, respectively). Similarly, the module Type1 fails to load because it does not exist. I'm not sure if the bug is in the modules themselves, or in Xconfigurator. The pex5 and xie modules are part of XFree86-compat-modules, and I don't think they're even needed / usable with XFree86 4.2. I'm not sure what, if anything, is supposed to supply Type1.so. The XF86Config-4 file generated during install, however, does include: Section "Module" <snip> Load "pex5" <snip> Load "xie" Load "Type1" It looks like either Xconfigurator / anaconda is buggy for putting those modules on the list, or the pex5 and xie modules are buggy for not loading....
PEX and XIE are now obsolete, and no longer shipped. The configuration tools will have to be appropriately modified to no longer reference PEX and XIE. Type1 module is included, however Red Hat Linux does not use the X server for serving fonts by default, we use xfs instead. The config tools should definitely not be loading the Type1 or freetype modules. Good catch.
Also, when I was cleaning out the unneeded modules from XF86Config-4, I noticed that pex5 was actually in there twice. It might be a good idea to do a sort -u on the list of modules before inserting them into the X config to prevent duplicates like that....
We're gonna get tons of problems when people upgrade since these modules were in peoples old configurations.
That may be so, however either way, XIE and PEX are no longer being shipped, nor supported. So the config tools need to ensure that they do not write out lines trying to load them. I'm not sure how to best handle the cases where existing config files have these options enabled. Even with the options enabled I believe, the server should start up ok, just spewing out some errors. Any suggestions on how to best handle this case? We'd have to handle both distro upgrades as well as just X upgrades, which implies that doing it in anaconda would not completely solve the problem. Perhaps in one of the XFree86 %post scripts.
Config tools both updated to not enable/use pex/xie now