Add debugging print to check aarch64 repl test

[gbaraldi]

No description provided.

[vtjnash]

      From worker 10:	tmpdir = "/cache/build/default-armageddon-7/julialang/julia-master/tmp/jl_18tn9T"
      From worker 10:	 pwd() = "/cache/build/default-armageddon-7/julialang/julia-master/tmp/jl_18tn9T"

[vtjnash]

      From worker 4:	stat(tmpdir) = StatStruct("." size: 4096 bytes device: 66307 inode: 54934720 mode: 0o040700 (drwx------) nlink: 2 uid: 1000 (juliaci) gid: 1000 (juliaci) rdev: 0 blksz: 4096 blocks: 8 mtime: 2022-09-28T08:11:37+0000 (just now) ctime: 2022-09-28T08:11:37+0000 (just now))
      From worker 4:	   stat(".") = StatStruct("." size: 4096 bytes device: 66307 inode: 54934720 mode: 0o040700 (drwx------) nlink: 2 uid: 1000 (juliaci) gid: 1000 (juliaci) rdev: 0 blksz: 4096 blocks: 8 mtime: 2022-09-28T08:11:37+0000 (just now) ctime: 2022-09-28T08:11:37+0000 (just now))
      From worker 4: 	 stat(pwd()) = StatStruct("." size: 4096 bytes device: 66307 inode: 54934720 mode: 0o040700 (drwx------) nlink: 2 uid: 1000 (juliaci) gid: 1000 (juliaci) rdev: 0 blksz: 4096 blocks: 8 mtime: 2022-09-28T08:11:37+0000 (just now) ctime: 2022-09-28T08:11:37+0000 (just now))

[vtjnash]

The first call to stat returns info on the wrong folder:
From worker 4: infoa = StatStruct("." size: 4096 bytes device: 66307 inode: 3700025 mode: 0o040775 (drwxrwxr-x) nlink: 15 uid: 1000 (juliaci) gid: 1000 (juliaci) rdev: 0 blksz: 4096 blocks: 8 mtime: 2022-09-28T16:36:41+0000 (1 minute ago) ctime: 2022-09-28T16:38:03+0000 (14 seconds ago))

note how old this folder was when last modified (1 minute), last accessed (14 s), and how many hard links it has (probably folders inside)

[gbaraldi]

So for some reason . doesn't update fast enough for the test? Because it does give the correct result after it.

[gbaraldi]

What were you trying to test in the last commit @vtjnash ?

[vtjnash]

I am trying to isolate which lines are important and which don't change the result

[gbaraldi]

From worker 4: realpath(".") = "/cache/build/default-armageddon-5/julialang/julia-master/julia-3e628d6290/share/julia/test"
From worker 4: realpath(tmpdir) = "/cache/build/default-armageddon-5/julialang/julia-master/tmp/jl_Rmtz2C"

[gbaraldi]

Yeah, the expansion of . didn't change quickly enough for this test. Which is odd.

[vtjnash]

A little part of me wonders if the path on aarch64 is just a little bit too long, and causes tmpdir_pwd to be printed several times here. e.g. the same as #44972. The order of these reads is a bit nonsensical. Thus the actual cd might not have happened yet, until the added @show cause a context switch.

(you can prove this theory by checking when get(ENV, "OLDPWD", "") gets set here)

[gbaraldi]

Yeah, it's as you expected. A context switch is what triggers it to change. Not sure what the fix for this is. The paths don't look too too long

[vtjnash]

Fixed by #47160