I tried installing Gentoo Linux from the 16MB .ISO, which I downloaded on 12APR2002. install.txt worked well, except the minor point that when it says "Oh, a new prompt!" I still had the regular hash mark prompt. (not sure if that is significant) Anyway, everything goes well until the "scripts/bootscrap.sh" which downloads and builds for a while and then dies. What occurs: - The terminal running the script stops. I can hit enter and scroll the text up, I can hit arrow keys and fill the screen with escape codes, etc. but nothing else. - All other terminals simply lock up completely, except for the terminal that is running "less" which is viewing "install.txt" -- Less continues to work perfectly, scrolling up and down and searching, etc., until I exit. After I exit from less, I get the same thing as the emerge terminal--it technically responds to keypresses but does nothing useful. - If I go to a new terminal window and hit enter to activate the terminal, I get the same thing. Nothing happens after I hit "enter" other than my characters being echod onto the screen, but no shell is started. - All terminals that are open and are actually running a copy of BASH simply lock up completely, not responding to anything at all in any way. It is as if the shell program ceases to exist and the system doesn't realize it. Additionally, CTRL-ALT-DEL does not work on any terminal. CTRL-C works only on the term running the script, cancelling the process officially which has already stopped anyway. After this, the term behaves like the others. Note that this does indeed stop--it isn't stuck on a step that happens to take a long time, unless any one single step can take over 20 mins on a fairly fast system without ever accessing the hard drive and while locking everything else up. Step to reproduce (if anyone else can reproduce this): 1) Follow instructions in /install.txt, including: # cd /usr/portage # scripts/bootstrap.sh Notes: - I am using the default options in /etc/make.conf for an Athlon/PPro which include the -march=pentiumpro option. At first I was using a few extra flags, but not since the first try at installing. - I have zeroed out the hard drive twice using 'dd' - I am using xfs for /, ext3 for /boot, and have 1GB swap. - This is probably not a hardware stability problem. The system is made of all top-quality parts and has successfully done "make world" in freeBSD several dozen times. It has also passed memtest86 without ECC enabled, and has ECC enabled for single and double bit errors. Everything superfluous is disabled (All IDE channels, USB, all COM ports, parallel port) and the BIOS settings are very conservative. - I have also tried doing an install without ever opening a second terminal - XFS is using recommended settings from install.txt, EXT2 boot partition is using defaults - Both the swap and main partition are below the 1024 cylinder boundry (which I did just to see if it would effect anything) - The last thing seen on the screen is: " checking for bzero... yes checking for calloc... yes checking for clock... " (FWIW, yes the clock is set) Hardware: - Single AthlonXP 1400MHz, well cooled - on Tyan TigerMP mobo with BIOS 1.04 (newest) - Single 1GB Corsair reg ECC memory module certified for this motherboard. - Matrox G400 video - Adaptec 29160 (64 bit version) SCSI adapter - Maxtor Atlas 10K III 18GB U160 hard drive (/ is a 7 GB partition) - Intel Pro100 NIC - Plextor 40X wide CD-ROM - Good power supply (400W PCPower & Cooling Silencer) - No sound, RAID, mouse, IDE devices, etc. - Not overclocked (In fact, I can't with this mobo) - When the system was running Windows 2000, I did a stability test in which I ran SETI@Home and 3DMark 2001 in a continuous loop for about 31 hours. No problems. - This system has had Redhat 7.2 W\ Ximian successfully run and installed for a Unix desktop demo I gave in class, so it probably isn't a hardware Linux incompatibility.
I'm having this exact same problem, just about. When I reach the installation instruction section code listing 17 and type source /etc/profile, I do not get a new prompt. Then, later, when I bootstrap, the machine hangs. In my case, I have absolutely no access to the machine. The screen isn't even just blank, it's inaccessible ... as though the machine's not even sending a blank screen to the monitor. No keystrokes work whatsoever. I've had the same results with several different make.conf flag combinations on a PIII 450, using ext3 for the file system.
Addendum: If it matters, I found why the shell prompt does not update when the instructions say: "# Ooh! A new prompt!" The /etc/profile shell script that the instructions say to "source" checks to see if $SHELL is "/bin/bash", but I have found that the $SHELL variable is always "/bin/sh" even if the shell is bash. "/bin/sh" is a symlink to "/bin/bash", so they are the same, but the variable is not the same so when "profile" checks the $SHELL variable, it sees the wrong one and never sets the prompt and whatever else it is supposed to set. Workaround: For users, prior to "source /etc/profile", type "SHELL = /bin/bash; export SHELL" For the developers: Have /etc/profile simply not even check the shell. It's not like csh or ksh is even installed at this point, and if someone does install it and it screws up the install, it's their own fault for not following the instructions. Alternatively, allow /bin/sh to be just as valid as /bin/bash
Update: I have successfully installed Gentoo using the stage 3 ISO, and while in Gentoo have successfully rebuilt everything without any problem. I suspect that the kernel used for the install may be the problem, and suggest that it be changed to a "safe" kernel, given that having the absolute latest kernel with high optimizations does not matter for a kernel that won't be used for more than the installation.
Yes, our current kernel's XFS dies with highmem; this will be fixed next release.