From Bugzilla Helper: User-Agent: Mozilla/4.7 [en] (WinNT; I) Description of problem: When attempting to install RedHat9 on a bochs2.1 emulator, the text installer crashes durring RPM installation. Version-Release number of selected component (if applicable): How reproducible: Always Steps to Reproduce: 1) Installed Bochs21. 2) configured startup script for network card, flat c.img file, and directly referanced D: as the the cdrom 3) configured startup script to boot to cdrom 4) configured the emulated NE2k netowrk card 5) installed from the CD in the cd-rom drive (RedHat9-disk1) 6) the emulated vgs and monitor interfaces aren't recognized, so the installed falls back to text mode 7) selected a typical Server configuration 8) after installing several RPMs the installer throws an error, and aborts the installatio Actual Results: Trackstack (most recient call last): File "/usr/bin/anaconda", line 739, in ? intf.run(id, dispatch, configFileData) File "/usr/lib/anaconda/text.py", line 459, in run dispatch.gotoNext() File "/usr/lib/anaconda/dispatch.py", line 157, in gotoNext self.moveStep() File "/usr/lib/anaconda/dispatch.py", line 255, in moveStep rc = apply(func, self.bindArgs(args)) File "/usr/lib/anaconda/packages.py", line 817, in doInstall problems = ts.run(cb.cb, 0) File "/usr/lib/anaconda/packages.py", line 359, in cb self.progress.setPackageScale(amount, total) File "/usr/lib/anaconda/textw/progress_text.py", line 55, in setPackageScale self.s.set(int(((amount * 1.0)/ total) * 100)) OverflowError: float too large to convert Kicking this into Debug and investigating the amount and total variables results in: > /usr/lib/anaconda/textw/process_text.py(55)setPackageScale() -> self.s.set(int(((amount * 1.0)/ total) * 100)) (Pdb) p total 2347440 (Pdb) p amount 2240824 (Pdb) p ((amount * 1.0)/ total) * 100 95.458201274579963 (Pdb) p int(((amount * 1.0)/ total) * 100) 95 (Pdb) c This then continues and returns an error, causing the install to be aborted. Expected Results: Succesful RPM installation for the selected modules. Additional info: After several attempts, there appears to be no pattern as to which RPM is being installed when anaconda crashes. This has typically occured arround 30% overall completion, so several RPMs are successfully installed before the crash.
With it being largely random, this sounds a lot like a "hardware" problem. Also, does it happen with FC1?
Since the exact same python statments run successfully in the debugger, and don't produce the Overflow Error... you're most likely correct. The bochs files do indicate extensive work was being done in the FPU for the 2.1 version, so it may take it a bit longer to mature. There was also a documented case in one of the MMX OP-codes that produced a race condition that could result in a floating point overflow. (Thought I'm not sure why the text version of the anaconda installer would be using the MMX op-cdes, but... it's possible.) Yes, the Fedora Core install produced the same results, only at 47% rather than around 30% completion. But it to aborted dussing the RPM install. Unlike the shrike install, it did not give me the option to debug the error, nor did it give me any actual error text. It just stopped processing and unmounted the filesystem. But it's most likely a simular problem with the bochs system. I'll continue to persue the "hardware issue", but at present this doesn't look like a RedHat problem. Thanks for the help.
As follow up to this chain... Bochs has released a 2.1.1 patched version, this does include an overload issue fix that appears to address this "hardware" problem, as FC1 does successfuly install when this version is used.