Closed tiran closed 1 month ago
We made a mistake during our testing last night. This change does not fix the infinitive recursion segfault in cmake. extlib_loguru.cmake
is also causing trouble. I don't have a good understanding of cmake. Whatever you are doing in https://github.com/DS4SD/deepsearch-glm/tree/main/cmake works and whatever docling-parse is doing differently fails for us.
@tiran ok, let me take a look.
@tiran I did not want to pollute the branch of your fork. If we want to do something similar as glm, this is done in this new PR (https://github.com/DS4SD/docling-parse/pull/44)
@tiran shall we close this? I think it is now all resolved with #44, right?
Yes, this PR is no longer needed with #44.
Our builds are failing with a segfault with cmake 3.26.5 on RHEL. The final phase of CMake is trapped in an infinitive recursion and eventually exhausts all stack trace. The segfault only occurs with
USE_SYSTEM_DEPS=ON
.Eventually we found the culprit in
extlib_cxxopts.cmake
. The lineadd_dependencies(${ext_name_cxxopts} cxxopts)
is causing the segfault.ext_name_cxxopts
iscxxopts
. I'm not entirely sure why it is failing. I think there is a variable name clash. Settingext_name_cxxopts
to a different value resolves the problem.