Boost 1.91 -> 1.92, Updated to mongodb-4.4.31, mongobleed patched upstream
There should be no impact from the backmp11 breaking changes in the boost update. I've grepped the source and the process_event, process_queued_events, and process_single_queued_event do not appear in the code. So at most a rebuild against boost-1.92 to pick up the new so version is all that should be required.
Note: patch for SConstruct required due to change to pkg_resources.
Thanks to Lone Wolf and loqs in my forum thread for help figuring out how to handle the changes to scons pkg_resources in SConstruct. We again have a fully functional and clean building mongodb44 package updated to the latest upstream source. Added mongodb-4.4.31-scons-pkg_resources.patch
Note: the mongobleed vulnerability in its zlib compressed header code that allowed unauthenticated users to leak uninitialized heap memory from the database was fixed in 4.4.30.
Also note, with the latest kernel and gcc-16, this package will no longer build on boxes with 4G of RAM with LTO enabled (ld terminated with signal 9 [killed], cc1plus invoked oom-killer). I did not try building without LTO, I just moved to another box.
Pinned Comments
drankinatty commented on 2026-08-21 02:46 (UTC) (edited on 2026-08-27 07:00 (UTC) by drankinatty)
Boost 1.91 -> 1.92, Updated to mongodb-4.4.31, mongobleed patched upstream
There should be no impact from the
backmp11breaking changes in the boost update. I've grepped the source and theprocess_event,process_queued_events, andprocess_single_queued_eventdo not appear in the code. So at most a rebuild against boost-1.92 to pick up the new so version is all that should be required.Note: patch for SConstruct required due to change to
pkg_resources.Thanks to Lone Wolf and loqs in my forum thread for help figuring out how to handle the changes to scons
pkg_resourcesin SConstruct. We again have a fully functional and clean building mongodb44 package updated to the latest upstream source. Added mongodb-4.4.31-scons-pkg_resources.patchNote: the mongobleed vulnerability in its zlib compressed header code that allowed unauthenticated users to leak uninitialized heap memory from the database was fixed in 4.4.30.
Also note, with the latest kernel and gcc-16, this package will no longer build on boxes with 4G of RAM with LTO enabled (ld terminated with signal 9 [killed], cc1plus invoked oom-killer). I did not try building without LTO, I just moved to another box.