This issue is related to a GCC bug(https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102124), which may cause wrong calculation, causing buf_eq_const() function give positive result on different buffers.

362 /* Constant-time compare of two buffers.  Returns 1 if buffers are equal,
363    and 0 if buffers differ.  */
364 static inline int
365 buf_eq_const(const void *_a, const void *_b, size_t len)
366 {
367   const byte *a = _a;
368   const byte *b = _b;
369   int ab, ba;
370   size_t i;
371 
372   /* Constant-time compare. */
373   for (i = 0, ab = 0, ba = 0; i < len; i++)
374     {
375       /* If a[i] != b[i], either ab or ba will be negative. */
376       ab |= a[i] - b[i];
377       ba |= b[i] - a[i];
378     }
379 
380   /* 'ab | ba' is negative when buffers are not equal. */
382   return (ab | ba) >= 0;
383 }

On line 382, I think it should be return (ab | ba) == 0;
Because if 2 buffers equal to each other, the difference between any value of the buffer should be 0.
If 2 buffers do not equal to each other, the difference between any value of the buffer should be non-zero.
Non-zero value is either positive or negative value.

Because the GCC's bug, 2 different buffer's diff value got a "positive" data, however, positive data still means 2 buffers are different, only if the diff value of 2 buffers are 0, means they are equal, in this case return (ab | ba) >= 0; is apparently wrong.