OpenEXR vs Typito

OpenEXR's recent history is dominated by security response. On 2026-08-05 three branches were tagged within two minutes — v3.4.14, v3.3.13 and v3.2.11, fixing 15, 15 and 10 CVEs from one fuzzing and audit campaign, all memory corruption reachable by opening a crafted .exr file. Two weeks later v3.4.15 and v3.3.14 fix two further memory-allocation issues, this time confined to the code that decodes the idmanifest attribute, with CVEs requested but not yet assigned. The 3.4 tag also carries a missing Windows export and reduced compiler warnings in example code.

The pattern is now established: findings arrive from a continuing fuzzing effort and are patched simultaneously across every supported stream, with the release notes written to tell integrators precisely which code paths are reachable. The narrowing scope is the useful signal — August's first batch spanned the C++ libraries, the command-line tools and the Python bindings, while this one touches only idmanifest decoding and explicitly states other code is unaffected even on files carrying the attribute. Both new tags are release candidates dated ahead of their tagging, and the 3.3 backport reuses the 3.4 note verbatim, down to naming v3.4.15 in its own body.

Expect the requested CVEs to be assigned identifiers and the two RCs to be promoted to final tags, with a matching 3.2 backport if the idmanifest code is present in that stream. Further attribute-parser findings are likely while the fuzzing campaign continues.