Test failures pre 3.8.0

[zanchey]

The tests are failing on the package builders in a few places.

I'm filing an issue for posterity and also to make sure they actually get fixed.

1. 32-bit builds fail in src/tests/common.rs:127:

assertion `left == right` failed
  left: "-9223372036854775808"
 right: "0"

@KamilaBorowska spotted this one as the argument to sprintf being long, which is 32 bits on a 32-bit platform rather than defined as 64-bit.

2. 32-bit builds fail in src/wutil/wcstod.rs:575:9

thread 'wutil::wcstod::test::tests' panicked at 'assertion failed: `(left == right)`
 left: `Ok(8.925500000000001e-18)`,
 right: `Ok(8.9255e-18)`'

@mqudsi notes this is the wrong comparison to be doing as floating point values should be compared by epsilon

I don't know why these two aren't picked up by the 32-bit CI build.

3. Unwritable home directories make the history tests fail.

---- tests::history::test_history_formats stdout ----
thread 'tests::history::test_history_formats' panicked at src/tests/history.rs:561:32:
Failed to get data directory

---- tests::history::test_history_merge stdout ----
thread 'tests::history::test_history_merge' panicked at src/tests/history.rs:395:9:
assertion failed: history_contains(&everything, text)

---- tests::history::test_history_races stdout ----
thread 'tests::history::test_history_races' panicked at src/tests/history.rs:336:5:
assertion `left == right` failed
  left: 1
 right: 1025

HOME is set to something unwritable in the Debian package builds. This wasn't a problem pre-cargo test, because tests/test_env.sh was run beforehand. The cargo tests should be equally hermetic. It probably means reimplementing test_env.sh in Rust, because sourcing it before running cargo test stuffs up any rustup-installed toolchains.

[KamilaBorowska]

Testing for equality in wcstod tests is actually correct, this is pointing a genuine problem. However, this is specifically caused by x87 floating point being pretty broken: rust-lang/rust#114479.

This problem affects Rust's standard library too. Rust's standard library has a pretty cursed workaround for this issue: https://github.com/rust-lang/rust/blob/master/library/core/src/num/dec2flt/fpu.rs.

Personally I would skip those tests for non-SSE2 x86 all(target_arch = "x86", not(target_feature = "sse2")).

Or alternatively, use str::parse rather than fast_float for those targets instead (as standard library has a workaround for this bug). That said, I don't think it's worth it, x87 is not IEEE compliant, so the results are going to be subtly wrong when doing mathematical operations anyway.

[mqudsi]

I pushed some updates that got tests working for me targeting i586-unknown-linux-gnu under Debian 12 i686. Using str::parse() instead of fast_float is not a one-liner given how we use the latter, so I'm just using a combination of my suggested test and your sse2 note.

The history_formats test failure is because we are including a source code test that relies on the cmake build/tests target being present. I'll fix that separately.

[mqudsi]

test_history_formats fixed in 476b360; everything should be passing now (EDIT: but that's locally, not under CI where the permissions issue exists separately!)

[mqudsi]

@zanchey after a35925b, tests appear to pass with the following under Debian 12 i686, which I assume does the trick for you?

set fish_home (mktemp -d)
sudo chown root:root $fish_home
env HOME=$fish_home cargo test --target i586-unknown-linux-gnu

(I don't actually run the last command as-is because cargo test actually calls rustup which uses $HOME for its own purposes; I actually run a bare cargo test --target i586-unknown-linux-gnu once which builds the test binary as target/debug/i586-unknown-linux-gnu/fish-<hexadecimal-string> then I run that binary with env HOME=$fish_home.)

[mqudsi]

@KamilaBorowska do you have any idea why the other 32-bit vs 64-bit errors didn't manifest under i686 but show up under i586? For example, I would expect the snprintf issue with %ld passed an i64 to show up regardless of which particular 32-bit flavor we're targeting.

[zanchey]

This is all working ok now, and I fixed the issues with Ubuntu's lto-wrapper (which breaks for some reason when linking the cargo tests) with f4a79cc.

[zanchey]

Unfortunately the 32-bit builds are failing on Debian again.

Testing file checks/math.fish ... Failure:

The CHECK on line 77 wants:
  1.316074

which failed to match line stdout:32:
  1.316075

Context:
  [...] from line 71 (stdout:29):
  1000000000000000
  100000000000000
  -1000000000000000
  1.316075 <= does not match CHECK on line 77: 1.316074
  4
  0
  1
  [...] from line 261 (stdout:77):
  2
  3
  0.841471
  -0.89001 <= does not match CHECK on line 283: -0.890009
  0.5
  4
  2

when running command:
  ../test/root/bin/fish checks/math.fish

and

---- tests::negative_precision_width stdout ----
thread 'tests::negative_precision_width' panicked at 'assertion failed: `(left == right)`
  left: `"78.900001   "`,
 right: `"78.900000   "`', printf/src/tests.rs:601:5

---- tests::test_float stdout ----
thread 'tests::test_float' panicked at 'assertion failed: `(left == right)`
  left: `"1.100001"`,
 right: `"1.100000"`', printf/src/tests.rs:359:5

---- tests::test_float_g stdout ----
thread 'tests::test_float_g' panicked at 'assertion failed: `(left == right)`
  left: `"1.23456789012346"`,
 right: `"1.23456789012345"`', printf/src/tests.rs:453:5

---- tests::test_locale stdout ----
thread 'tests::test_locale' panicked at 'assertion failed: `(left == right)`
  left: `"-46,380001"`,
 right: `"-46,380000"`', printf/src/tests.rs:702:9