[2.3.0~beta1 regression] Long columns are misaligned

[kit-ty-kate]

$ ./opam-2.3.0-beta1-arm64-linux upgrade
The following actions will be performed:
=== recompile 9 packages
  ↻ dns-client-miou-unix     9.0.1
          [uses happy-eyeballs, happy-eyeballs-miou-unix]
  ↻ h1                       1.0.0                       [uses angstrom]
  ↻ h2                       0.13.0                      [uses angstrom]
  ↻ hpack                    0.13.0                      [uses angstrom]
  ↻ httpcats                 0.0.1
          [uses happy-eyeballs-miou-unix]
  ↻ kit-ty-kate-platform     1                           [uses ocp-index]
  ↻ lambda-term              3.3.2                       [uses zed]
  ↻ uri                      4.4.0                       [uses angstrom]
  ↻ zed                      3.2.3                       [uses uucp, uuseg]
=== upgrade 7 packages
  ↗ angstrom                 0.16.0 to 0.16.1
  ↗ caqti                    1.9.0 to 2.1.2
  ↗ happy-eyeballs           1.2.0 to 1.2.2
  ↗ happy-eyeballs-miou-unix 1.2.1 to 1.2.2
  ↗ ocp-index                1.3.6 to 1.3.6.1~alpha-repo
  ↗ uucp                     15.1.0 to 16.0.0
  ↗ uuseg                    15.1.0 to 16.0.0
=== install 3 packages
  ∗ dune-private-libs        3.16.0
          [required by dune-site]
  ∗ dune-site                3.16.0                      [required by caqti]
  ∗ lwt-dllist               1.0.1                       [required by caqti]

Proceed with ↻ 9 recompilations, ↗ 7 upgrades and ∗ 3 installations? [y/n] n
$ ./opam-2.3.0-alpha1-arm64-linux upgrade
The following actions will be performed:
=== recompile 9 packages
  ↻ dns-client-miou-unix     9.0.1                       [uses happy-eyeballs, happy-eyeballs-miou-unix]
  ↻ h1                       1.0.0                       [uses angstrom]
  ↻ h2                       0.13.0                      [uses angstrom]
  ↻ hpack                    0.13.0                      [uses angstrom]
  ↻ httpcats                 0.0.1                       [uses happy-eyeballs-miou-unix]
  ↻ kit-ty-kate-platform     1                           [uses ocp-index]
  ↻ lambda-term              3.3.2                       [uses zed]
  ↻ uri                      4.4.0                       [uses angstrom]
  ↻ zed                      3.2.3                       [uses uucp, uuseg]
=== upgrade 7 packages
  ↗ angstrom                 0.16.0 to 0.16.1
  ↗ caqti                    1.9.0 to 2.1.2
  ↗ happy-eyeballs           1.2.0 to 1.2.2
  ↗ happy-eyeballs-miou-unix 1.2.1 to 1.2.2
  ↗ ocp-index                1.3.6 to 1.3.6.1~alpha-repo
  ↗ uucp                     15.1.0 to 16.0.0
  ↗ uuseg                    15.1.0 to 16.0.0
=== install 3 packages
  ∗ dune-private-libs        3.16.0                      [required by dune-site]
  ∗ dune-site                3.16.0                      [required by caqti]
  ∗ lwt-dllist               1.0.1                       [required by caqti]

Proceed with ↻ 9 recompilations, ↗ 7 upgrades and ∗ 3 installations? [y/n] n

Funnily enough master (built with glibc) exhibits the same behaviour as 2.3.0~alpha1 so my guess is that this somehow comes from #6237, most likely the upgrade to Alpine 3.20. Maybe a bug in musl?

EDIT: Nevermind i built the wrong branch, i'm able to reproduce with master with glibc also so it's not a musl issue.

[kit-ty-kate]

I was able to bisect quickly and found the origin in #6230. Apparently tput cols says something different depending on whether it's called via execve or system or something like that. I'll find a fix tomorrow

[kit-ty-kate]

Something feels fishy in the standard library:

# let ic, _ = Unix.open_process_args "/usr/bin/tput" [|"/usr/bin/tput"; "cols"|];;
val ic : in_channel = <abstr>
# input_line ic;;
- : string = "107"
# let ic, _, _ = Unix.open_process_args_full "/usr/bin/tput" [|"/usr/bin/tput"; "cols"|] (Unix.environment ());;
val ic : in_channel = <abstr>
# input_line ic;;
- : string = "80"

[kit-ty-kate]

ok, i got to the bottom of this. It looks like you can't get the proper result from tput cols while redirecting all 3 stdout, stdin and stderr. At least one has to be left unredirected.

$ tput cols
215
$ tput cols > /tmp/stdout << EOF
EOF
$ cat /tmp/stdout
215
$ tput cols > /tmp/stdout 2> /tmp/stderr
$ cat /tmp/stdout
215
$ tput cols > /tmp/stdout 2> /tmp/stderr << EOF
EOF
$ cat /tmp/stdout
80

I'll try to think at a fix tomorrow

[dbuenzli]

I guess the idea is that you need one that isatty no ?

[dbuenzli]

Btw. I was curious on how I solved that problem in down as I knew I was not using tput. It's quite ugly :-) (get cursor position, move a large amount to the right this gets clamped if larger than the window size, get cursor position, and move back to initial position).

[kit-ty-kate]

My first thought on how to fix this so far is to bind ioctl(STDOUT_FILENO, TIOCGWINSZ, &win), it's easy to add in our codebase and seems reasonably portable (despite not being POSIX defined)

[dbuenzli]

Looks cleaner.

I'm not sure why I didn't do that in down but perhaps the thinking was let's just assume we have ANSI/VT100 terminal. If that turned out to make it work out of the box with @nojb's recent Windows Console integration, I'm less embarrassed :–)