Create a sync destination's parent, and retry a lost create-script

[titouanmathis]

Two bugs found by a real first deploy of www.alloayiti.fr, neither of them covered by the tests.

rsync never creates the destination's parent

rsync web/index.php → server:~/www/preview-462--app/web/index.php
rsync: [Receiver] change_dir#3 ".../web" failed: No such file or directory (2)

rsync makes the destination directory itself, but not its parent. So a single file whose parent is missing fails, and whether a deploy worked depended on the order of --sync: a project listing web/index.php before the web/wp that would have created web/ never got past the first file. Scalian only escaped it because its path order happened to create web/ first.

The parent is now created by --rsync-path, which runs in the remote shell and therefore works with any rsync version. Not --mkpath: that needs rsync 3.2.3 at both ends, and a runner older than that would fail on the option itself.

A failed first deploy lost its create-script for good

That failure exposed a worse one. The create-script was keyed on the project directory being absent — but the failed deploy had already cloned, so every later deploy treated it as an existing project and skipped the create-script. The environment stayed unseeded permanently, with no recovery but to destroy it.

On the server it left exactly that:

dir exists: yes   .git: yes   vendor: yes
.env: MISSING     web: no

.env was written by the create-script, so no retry would ever produce it.

It is now keyed on a marker written only after the script succeeds, so a create-script that fails runs again next time. Three cases, all tested:

State on the server	Create-script
.trafic-created present	skipped
.trafic-cloned present, .trafic-created absent	runs again — a first deploy of this version that failed before it
neither, but the directory exists	skipped and recorded — predates the markers

That last row is deliberate: re-running a seed against a database that has been live for weeks is worse than skipping a step that was probably already done. .trafic-cloned is what separates it from a genuine retry.

Tests

482 pass. Five new tests for the create-script decision (including the two cases above, which the old logic got wrong) and four for parentOf and the --rsync-path argument. CLI coverage 97.16% statements / 93.37% branches, above the gate added in #52.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 82.30%. Comparing base (0b6831d) to head (85b65e1).

Additional details and impacted files

@@            Coverage Diff             @@
##             main      #53      +/-   ##
==========================================
+ Coverage   82.12%   82.30%   +0.18%     
==========================================
  Files          35       35              
  Lines        1466     1481      +15     
  Branches      374      375       +1     
==========================================
+ Hits         1204     1219      +15     
  Misses        219      219              
  Partials       43       43              

Flag	Coverage Δ
integration	67.54% <100.00%> (+1.80%)	⬆️
unit	82.10% <100.00%> (+0.18%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.