rustc fails w/ io_error on F18, i686

[itdaniher]

I'm using a stage2 build cross-compiled on a x86_64 F19 machine. I've also tried using the stage0 snapshots at

http://static.rust-lang.org/stage0-snapshots/rust-stage0-2013-11-28-859c3ba-linux-i386-072e638a6a11f7d00cf2c7b487162d0d2e3b5036.tar.bz
http://static.rust-lang.org/stage0-snapshots/rust-stage0-2013-12-07-49b751d-linux-i386-e3c11032b19a74b86b5b6f271ee7457ce0b00d48.tar.bz2

All fail with the following error:

task 'rustc' failed at 'Unhandled condition: io_error: io::IoError{kind: OtherIoError, desc: "no such file or directory", detail: None}', /home/it/projects/rust/src/libstd/condition.rs:131
error: internal compiler error: unexpected failure

I jumped ran in gdb using:

gdb --args ./glibc-2.17-4.fc19.i686/lib/ld-linux.so.2 --library-path $PWD/glibc-2.17-4.fc19.i686/lib:$PWD/stage2:$PWD/stage2/lib/rustc/i686-unknown-linux-gnu ./stage2/bin/rustc -L ./stage2/lib/ -L ./stage2/lib/rustc/ stage2/main.rs

and narrowed the error down to:

task 'rustc' failed at 'assertion failed: res as int != 0', /home/it/projects/rust/src/libextra/flate.rs:84

It's worth noting that the glibc path hack is not necessary for the snapshots, but they do require manually passing the library search flag via rustc -L.

The original error resembles that described at #10754 and potentially addressed in part by #10863.

[alexcrichton]

I'm not sure that the unhandled condition and the assertion error are quite related, but the assertion error is somehow probably related to deflation of metadata hitting an error. When you say that you're using the stage0 snapshots, what do you mean by that? Are the steps to produce this simply to check out the repo and run make or are the some more things in play?

I think that the -L directories your passing to gdb may be containing different versions of the same library which may be causing flate to go haywire, but the original unhandled condition problem is disturbing. Could you get a backtrace for when the unhandle condition error is thrown?

[itdaniher]

When you say that you're using the stage0 snapshots

I'm using the stage0 bootstrap snapshots from the links above. They're stripped of debug info, which makes life a little difficult.

I've also tried cross-compiling, but I have a different glibc on my F18 and F19 machines, requiring a bit of path hackery.

Could you get a backtrace for when the unhandle condition error is thrown?

Not past what I've been able to produce - stripped info, etc.

Are the steps to produce this simply to check out the repo and run make or are the some more things in play?

The machine I'm trying to run rustc on has a whopping 256MB of RAM, so compiling rust in situ is a bit of a stretch.

[treeman]

Triage bump.

@itdaniher status?

[itdaniher]

My creative solution was to abandon trying to run rustc on that platform. If nobody's reproduced in the interim, bug is probably on my end. Closing issue.