zentralopensource / zentral / 30468671944
90%

DEFAULT BRANCH: main
Ran
Jobs 1
Files 1016
Run time 2min
Badge
Embed ▾
README BADGES
x

If you need to use a raster PNG badge, change the '.svg' to '.png' in the link

Markdown

Textile

RDoc

HTML

Rst
coverage: 89.54%. Remained the same
30468671944

push

github

np5 
Hand a JSON tree its hash back instead of writing it into the tree

prepare_commit_tree() marked every subtree it walked with an mt_hash key,
and cleanup_commit_tree() took the markers back out afterwards. In a JSON
subtree that in-band marker cannot be told from a key a client happens to
call mt_hash, so the key was stripped before hashing — and it went to the
column stripped as well, which left it the one value the JSON path still
lost now that the empty ones are kept.

prepare_commit_tree() returns the digest. A model tree still gets the key,
because that is how commit() walks a tree that is already prepared and how
the object takes its mt_hash, and a JSON tree gets nothing written into it.
So mt_hash is data like any other key there, hashed and stored, and it
cannot claim the identity of anything: the digest is computed over it, not
taken from it.

Both call sites that cleaned markers out of a JSON value go away with the
marker, which leaves cleanup_commit_tree() unused — dropped. What reaches
the column is now a plain copy of what was hashed.

No committed column holds an mt_hash key, since it was always stripped on
the way in, so no committed row changes hash either: of the 43,964 rows the
deployed algorithm can commit, 0 are orphaned over 200k fuzzed payloads.
Coverage of the module stays at 100%, statements and branches.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

49835 of 55657 relevant lines covered (89.54%)

0.9 hits per line

Coverage Regressions

Lines Coverage ∆ File
51
89.39
 -0.01% zentral/contrib/inventory/models.py
Jobs
ID Job ID Ran Files Coverage
1 30468671944.1 1016
89.54
 GitHub Action Run
Source Files on build 30468671944