In order to work well, a jamming signal must be well-correlated with the signal it is trying to jam. In the case of sine-wave modulated signals, it's easy to detect the signal one wishes to jam, to infer that it's narrowband (and, hence, riding on a real or suppressed sine-wave carrier), to infer its bandwidth, and from there to choose a suitable jamming signal.
To jam a CDMA spread spectrum signal you'd have to find it, find its PN sequence, synchronize to its PN sequence, and then in most cases adjust your synchronization to match the time offset between your reception of the signal and the receiver's, then you'd have something to send -- but if I were trying to receive the signal, I could defeat you by having multiple receivers, located more than a wavelength of the chipping rate away from each other, each trying to receive the signal. One receiver may get the jamming signal and the intended signal on top of each other, but unless your jamming signal is co-located with the intended transmission, other receivers would be able to pick out the good from the bad.
But keep in mind that civilian CDMA is intended to minimize unintentional jamming, and does pretty well. It's not intended to defeat malicious jamming.
I suspect that there are smart military types out there that know a lot more than I do about jamming CDMA -- I'm just making this up as I go from first principles. I also suspect that military systems will use something more complicated than civilian CDMA, for just that reason.