It turns out that according to valgrind and perf, it's also slightly faster.
Details
Details
Diff Detail
Diff Detail
Event Timeline
Comment Actions
Concerning the performance aspect, when counting the number of instructions needed to compile the sqlite amalgametion, using
valgrind --tool=callgrind ./bin/clang -c sqlite3.c -o/dev/null -w
I get 5,872,876,067 instructions after the patch, and 5,879,411,049 before.