On Tue, 15 Jul 2003, Mark Vojkovich wrote: >> Has anyone investigated the root cause of why there is no fixed font in >> the first place -- like why the font server isn't running? Was it never >> started (a vendor-specific configuration problem), or did it crash (a >> robustness problem with xfs)? >> > > That's a good question. I always assumed it was a configuration >problem since all the reports came from specific versions of >Red Hat.
If someone can take this beyond assumptions to actual proof of it being a Red Hat packaging or configuration flaw, I'd be glad to investigate it. Otherwise I don't consider it to be Red Hat specific. It very well however could be caused by a specific font or somesuch, and if that's the case, reproducing it on another distro would be a matter of installing the faulty font. > Regardless, I think there is a fundamental robustness >problem with the configuration used by these vendors. There >is no reason to not have BOTH a font server and explicit >font paths in the XF86Config. That way the X-server still >runs if there isn't a font server, but if the font server >is listed first, it will get used by default if it is >running. Font paths are created dynamically as optional font packages are installed. When a font package is installed, it's font path is added to the xfs configuration, and xfs is sent a SIGUSR1 to reload it's config file. The work required to change the tools to add fontpaths to both xfs and the XFree86 config file are nontrivial and would be a major change to incorporate into the distribution for no real benefit other than speculative. Such a change at this point would more likely than not, introduce more regression in the font subsystems than any perceived benefits it might bring. It would also require us to officially support 3 completely different methods of installing and configuring fonts. With the whole world moving to Xft/fontconfig, and moving away from core fonts, it is definitely not something of a priority to direct significant engineering resources to changing the way core fonts are installed and handled, and breaking a lot of compatibility of existing internal and external packages in the process for no real gains, increasing technical support issues and bug reports. I just see xfs and core font problems as being hot potatoes that nobody is that interested in really fixing, in particular if it can't easily be reproduced on $THEIR_OS_DISTRIBUTION, and is more interested in casting blame at $VENDOR instead. I'd prefer to just compile a few fixed fonts into the X server directly, and then get rid of xfs completely, but that's not likely something we can do for a year or more yet. The number of bug reports I've seen incoming each release surrounding xfs and core fonts has significantly decreased, partly due to upstream bug fixes that have went into xfs, mkfontdir, etc. and partly due to changes I've made to our packaging and other infrastructure as well. Feel free however to pinpoint specific problems in our font installation/configuration infrastructure, or specific problems in our font handling methodology to me, and I will certainly investigate making improvements if possible, and if the scope of the benefits of potential changes can outweigh any engineering resource costs associated with making such changes without massive distro-wide font mayhem catastrophes. -- Mike A. Harris _______________________________________________ XFree86 mailing list [EMAIL PROTECTED] http://XFree86.Org/mailman/listinfo/xfree86
