On Wednesday 04 June 2008 19:46:30 Paul Mundt wrote: > On Wed, Jun 04, 2008 at 12:36:11PM -0500, Rob Landley wrote: > > Actually if you ever need to diagnose early boot stuff on _any_ platform, > > you do need a console. But it can be serial or netconsole, as long as > > that works... > > Except for the minor fact that most early boot debugging happens long > before the console subsystem is even available.. Isn't that why CONFIG_EARLY_PRINTK was written? (And I mentioned Linus's hack using the RTC to see how far the _really_ early stuff got.) > At the risk of perpetuating the stupidity of this thread.. I plead the fifth. > If you ship a > device to a customer expecting them to debug it for you, you are likewise > not likely to be very commercially successful, either. There are such things as field servicable devices where companies either send people out into the wild or get hardware brought in for service. However, looking at the message you're replying to, I was talking about during development. (Remember how the early linksys boxes didn't have the serial port physically wired up to the outside, but if you busted out a soldering iron you got a shell prompt on it without even installing new software? They didn't change it after they got it working because they didn't want to re-validate their image? Yeah, that kind of stuff gets shipped.) Sure if the device in the field doesn't boot, it's cheaper to just replace it, unless you need to get data off the sucker (in which case they bring it in). But the defective units returned to the factory sometimes get diagnosed so they can reduce the future return rate (or resell 'em used if it was pilot error), so once again it helps if it's possible to service 'em after the fact. Depends on the company. > Devices are not shipping with consoles, period. Earlier I was trying to distinguish between /dev/console (no controlling tty), virtual terminals (tied to old VGA hardware although possibly usable through the framebuffer, I'm unclear on this), the tty layer (which is what I initially thought the patch was aimed at, but it seems to be the vga VT stuff), and having a bitmapped display (may be GUI only). Four separate things, I've lost track of which we're talking about here. When you imply it's stupid to think anyone will ever have a console on an embedded system (because we all know the embedded world is far more uniform than crazy diverse things like the desktop space), which of these are _you_ referring to? My cable modem has a serial port that gives me a login prompt if I plug into it, and I watched the guy at sprint bring one up last time I broke my cell phone so he could get my number list off it. That kind of console may be useful out in the real world, and that's the kind of console I was talking about in the message you replied to. I've learned that "I don't do that, therefore nobody ever will" is not always the world's greatest assumption. Somebody out there may want a console, and they may not want the overhead of internationalizing it because although their customers speak mandarin, their field service people do not. Somebody out there may also want to do something I consider a bad idea. (Compared to shipping a device with Windows CE on it, any quibble I have about Linux configuration is a rounding error.) > If you disagree with this, you've > obviously never shipped a device. Since I haven't shipped only _one_ device, and since I used to own a tuxscreen phone which may actually meet _all_ the above definitions of "console", I guess it's ok for me to disagree with this? Rob -- "One of my most productive days was throwing away 1000 lines of code." - Ken Thompson. -- To unsubscribe from this list: send the line "unsubscribe linux-embedded" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
