Autocorrelation remained the workhorse technique for measuring ultrashort laser pulses from the 1960s to the 1980s. It gave a rough idea of how long the pulse was, which was good enough for many applications.
Unfortunately, using the event to measure itself isn't good enough. It's only as short as the event; it's not shorter. This means that an autocorrelation measurement of a pulse won't tell us about any detailed structure in the pulse intensity vs. time. And of course, it gives no information at all about the phase.
Notice how details of the pulse intensity are smeared out in its autocorrelation.
This information is lost forever.
We proved this way back in 1986.
Another version of autocorrelation, called interferometric autocorrelation, gives some additional information--but not enough to determine the spectral phase or the intensity or phase vs. time or frequency.