Le 11/07/2018 à 10:48, Frederic Lecaille a écrit :
This is a patch to fix a very minor bug. But at least it helps in making
reg tests with the address sanitizer passes.

Hi Fred,

There is no leak here because realloc is used (using my_realloc2 wrapper). And in fact, it is important to call it twice. The first time, with the default value for global.max_syslog_len, to allow logging during the configuration parsing. The second time, after the configuration parsing to eventually resize the log buffers using the right value for global.max_syslog_len (deduced from the configuration).

Note that only the main thread will do the first allocation. The others will only do the second one because they are created after the configuration parsing.

Christopher Faulet

Reply via email to