From Bugzilla Helper: User-Agent: Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; rv:0.9.4) Gecko/20011019 Netscape6/6.2 Description of problem: crash occurs near the end of installing packages. How many floppies should I need for the crash dump? It kept telling me to insert a floppy, but it never wrote very much. Trace back is: File "/usr/bin/anaconda", line 620, in ? inf.run(id, dispatch, configFileDqta) File "/usr/lib/anaconda/text.ph", line 408, in run dispatch.gotoNext() I am trying to install a "small" system on a HP VE 5/75 to be a home router. It has 2 3c509 ethernet cards, 32M of RAM, and a 502M disk drive. Version-Release number of selected component (if applicable): How reproducible: Always Steps to Reproduce: 1. The autopartitioning refused to use such a "small" drive. (Linux has bloated. I remember when 0.5G was a big drive.) I used druid to use 21M for /boot, 64M for swap, and the rest for /. 2. Unfortunately, I don't know how to get a list of the packages I was trying to install. I turned off the X and Gnome/KDI stuff. I turned on the routing and emacs package groups. Actual Results: It fails during the emacs package at 332M of 361M. The / partition is about 450M. Expected Results: The package should have installed. Is there a big intermediate uncompressed file that won't fit? Maybe it is time to learn jove. No way am I going to learn vi. Additional info:
Trying it again with out emacs, I noticed that / has only 418M. I forgot that I had cranked up the swap space from 32M to 64M.
Yep. I think the problem was not enough room for intermediate files. The fix is to generate a decent error message. Instead of telling me to file a bug report, tell me that there is not enough room to install a package and the name of the package. Give me the option of continuing the install without said package. It would have been a lot easier adding the jed package after the install rather then filling out the druid form over and over. BTW, autopartition should not refuse to partition the disk. Instead, it should pop up a warning that my / partition is only <whatever> and that <?> is the recommended minimum. Why does anaconda need several 10s of millions of bytes of intermediate storage? The next install failed with only 3M left to go while installing jed using just the group level selections except for Applicaiton/Editors. The next iteration I went through all of the packages and removed the ISDN package and the stuff that needed PYTHON. A df -k of / ends up less than half full: 190279 used of 415361. (My first Linux install was a Slackware distribution from a few boxes of floppies. Back then, I think a 200M drive was considered too small and 500M was considered bleeding edge big.)
What type of install is this on?
custom
If this fails at emacs install, check out bug # 57842
What install medium are you using (cdrom, ftp, etc)?
cdrom I haven't looked at 57842, but this problem is not unique to emacs. emacs may tend to trip it because it is a big package.
It sounds possible that the cdrom is damaged somehow. When the crash happens, press <Ctrl><Alt><F4>. Do you see any messages that look related to the cdrom drive? Specifically, look for messages that say something about sectors or seek errors. For example: <4>end_request: I/O error, dev 16:00 (hdc), sector 936056 <4>hdc: cdrom_decode_status: status=0x51 { DriveReady SeekComplete Error } <4>hdc: cdrom_decode_status: error=0x34 <4>hdc: cdrom_decode_status: status=0x51 { DriveReady SeekComplete Error } <4>hdc: cdrom_decode_status: error=0x34 <4>hdc: cdrom_decode_status: status=0x51 { DriveReady SeekComplete Error } <4>hdc: cdrom_decode_status: error=0x34 Do you see anything like this?
Sorry about the delay, but I would have to dig that computer out of a pile and reinstall to look at the log. But then it occurred to me that I have since built a new computer and loaded everything on it from the same CD-ROMs with no problem. In fact, I did an "everything" install a couple of times without problems. So I think the media itself is OK.
Maybe a bad CD / media interaction. Without more information, that's the assumption I'm going on (because if it were really something else, I'd expect to see it in my testing or a lot more in bugzilla)