malloc(): invalid size (unsorted) about build-recorder HOT 3 CLOSED

After some research, it seems we've landed ourselves the most notoriously evil bug in the history of manual memory management, memory corruption.

Apparently, somewhere, something illegally writes on memory it shouldn't. Memory which belongs to the mighty one, malloc(3) itself. Little do i know about solving these sort of bugs besides reverse engineering them so i call upon the knowledge of the ancients for guidance. For thee are the only ones capable of taming this beast.

We've also sinned before, with multiple memory leaks throughout the code which i will address in a future pull request in an attempt to redeem my soul and achieve enlightenment. May thy forgive us.

from build-recorder.

Through some gdb magic I've discovered that the bug lies in record.c:get_cmdline(3) and am now working on it.

from build-recorder.

Aaaaaahhhhh, get_cmdline was allocating a buffer ret and then instantly started strcat(3)ing into it, without ever emptying the string first, which caused undefined behavior(that is, we couldn't possibly know where strcat(3) would start writing from or if it would even ever write at all).

Even worse strcat(3) ended up overwriting past-the-end of the allocated buffer, resulting in the malloc(3) failures that we witnessed. Such a trivial bug brought forth so much pain, makes you think.

from build-recorder.