[ericof]

Adds the Plone 1.0.6 image — the eleventh and last series. The matrix is complete: every release from 1.0 to 4.2 now ships.

Closes #197

Plone 1.0.6 with CMF 1.3.3 on Zope 2.7.8 / Python 2.3.7 — the same stack 2.0 uses. No new shared scripts.

This is the only image assembled from more than one artifact, and the only one that patches upstream source.

Three sources for one image

No single 1.0.6 artifact carries a working Plone. CMFPlone/INSTALL.txt asks for CMFPlone, DCWorkflow and Formulator plus CMF 1.3+, and the 1.0.6 tarball holds only the first.

Source	Supplies
dist.plone.org/archive/Plone-1.0.6.tar.gz	CMFPlone — and nothing else
dist.plone.org/archive/CMFPlone1.0.5.tar.gz	DCWorkflow, Formulator
old.zope.dev/.../CMF-1.3.3.tar.gz	CMFCore, CMFDefault, CMFTopic, CMFCalendar

Neither DCWorkflow nor Formulator is separately downloadable any more — old.zope.dev 404s for both — so the 1.0.5 release bundle is the only surviving source for the two companions of a 1.0.6 CMFPlone.

The two patches

Both are consequences of running a Zope 2.6-era Plone on 2.7. They live in 1.0/patches/ as unified diffs applied with patch -p1 --forward --batch, never as in-line sed. --forward makes an already-applied patch a hard failure rather than a silent re-application, and the build stops if either is rejected.

0001-isExpired-datetime-syntaxerror.patch widens an exception handler. Zope 2.6's DateTime raised IndexError on an unparseable string, which is what the handler was written against; 2.7's raises DateTime.SyntaxError, and CMF hands it the literal string 'None' whenever no expiry is set — so the exception escaped and every page render 500'd.

0002-login_form-tal-nesting.patch splits HTML close-tags inside a <script> block across a string concatenation. Zope 2.6's parser tolerated them inline; 2.7's HTMLTALParser raises NestingError and the template fails to compile.

The shared gate absorbed a third era, again unchanged

Plone 1.0 predates the <meta name="generator"> tag that every other series is asserted against. shared/smoke-test.sh fell back to a plone.css resource reference on its own:

OK: site root renders Plone markup via plone_javascripts/plone.css
    (pre-2.0: no generator tag)   (21167 bytes)

That is the third era-specific behaviour it has handled without modification, after discovering the two different site factories during the 2.1 and 2.5 ports. Nothing version-specific was added to the gate for any of the four tarball-era ports.

Two claims corrected rather than left to rot

legacy/2.0/README.md opened with "The oldest image in the matrix" — accurate until this commit, and now contradicted by 1.0/README.md's own first paragraph. It states what remains true: 2.0 is the oldest image that runs upstream source unmodified.

legacy/README.md's "one series lands per pull request; the remaining one follows these" is replaced by a completion statement noting why 1.0 is the sole patched image.

Verification

• make lint — 12 Dockerfiles, 11 shell scripts, workflow YAML
• make -n on all four 1.0 targets, including the test-demo-1.0 case the GNU Make 3.81 stem rule would otherwise route to test-%
• SMOKE_CREATE_SITE=1 make test-1.0 — 6/6, site root 21167 bytes
• make test-demo-1.0 — 6/6, build-time password confirmed dead after the ADMIN_PASSWORD reset

Read out of the built image, not from the ARGs: Products/CMFPlone/version.txt → 1.0.6, interpreter → 2.3.7, simplejson → 1.9.3, both patch markers present in the patched files, and Products/ holds exactly the seven products the three sources are meant to supply — CMFPlone, DCWorkflow, Formulator, CMFCore, CMFDefault, CMFTopic, CMFCalendar.

Both builds were cache-hit from the plone-legacy checkout, so the cold build is exercised for the first time in this PR's CI run.