happyz
/
llama.cpp
mirror of https://github.com/ggerganov/llama.cpp.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

server: workaround new chat parser regression

This commit is contained in:
James O'Leary 2026-03-18 12:30:44 -07:00
parent 5744d7ec43
commit 841e09eaf4
1 changed files with 9 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

10
tools/server/server-task.h
View File

@ -378,7 +378,15 @@ struct server_task_result_cmpl_final : server_task_result {
virtual void update(task_result_state & state) override { virtual void update(task_result_state & state) override {
is_updated = true; is_updated = true;
oaicompat_msg = state.update_chat_msg(content, false, oaicompat_msg_diffs); // common_chat_parse can throw std::runtime_error when the PEG
// parser fails to match the final output (is_partial=false).
// During streaming, the same text was parsed with is_partial=true
// and succeeded. Only this final parse can throw.
try {
oaicompat_msg = state.update_chat_msg(content, false, oaicompat_msg_diffs);
} catch (const std::exception & e) {
SRV_WRN("Final chat parse failed, using last partial result: %s\n", e.what());
}
oai_resp_id = state.oai_resp_id; oai_resp_id = state.oai_resp_id;
oai_resp_reasoning_id = state.oai_resp_reasoning_id; oai_resp_reasoning_id = state.oai_resp_reasoning_id;