I'VE MADE MY SHARE OF ERRORS, SOME OF WHICH ARE DETECTED ONLY AFTER PUBLICATION.  HERE IS A LIST OF SOME THAT I AM AWARE OF; PLEASE LET ME KNOW IF YOU DETECT OTHERS.

Thanks to Gennady Shmonin for pointing this out...
In the proof of Theorem 11,d_i=p_i=1 should be d_i = p_i = y_i 
Thanks to Marko Bertogna for pointing this out...
 Equation 3 is incorrect.  The second term in the min  -- A_k+D_k-C_k -- implicitly assumes that the job missing its deadline executes for C_k time units, whereas it actually executes for strictly less than C_k.  Hence this second term should be -- A_k+D_k-(C_k-\epsilon); for task systems with integer parameters, \epsilon can be taken to be equal to one.
(A similar modification needs to be made for the definition of I_2(\tau) -- Equation 5 in the paper)

Thanks to Marko Bertogna for pointing this out...

Claim 5 in this paper states:
This statement is incorrect (although the error is not relevant to the remainder of the paper).  The correct statement of this claim is as follows:
The corrected claim appears in the journalization of this paper (Sanjoy Baruah, Vincenzo Bonifaci, Alberto Marchetti-Spaccamela, and Sebastian Stiller. Improved multiprocessor global schedulability analysis. Real-Time Systems 46 (1), 2010.)

Figure 1 in this paper incorrectly claims that the non-cyclic GMF task model is a generalization of the "traditional" GMF model.  This is incorrect; a corrected version of this figure appears in (Sanjoy Baruah. The non-cyclic recurring real-time task model. Proceedings of the IEEE Real-Time Systems Symposium (RTSS), San Diego, CA. December 2010. IEEE Computer Society Press.)

Thanks to Haitao Zhu for pointing this out....
In the appendix, it is claimed (bullet item 4): 
For a deadline miss to occur, it is necessary that {\cal L }> {\cal U}; hence {\cal L} \leq {\cal U} is a sufficient schedulability condition.
This is incorrect; the correct statement is
For a deadline miss to occur, it is necessary that {\cal L }< {\cal U}; hence {\cal L} \geq {\cal U} is a sufficient schedulability condition.
Thanks to Gurulingesh Raravi for informing me of some significant errors in this paper. A corrected (and far more general) version of the results appear in (Andreas Wiese, Vincenzo Bonifaci, and Sanjoy Baruah. Partitioned EDF scheduling on a few types of unrelated multiprocessors. This is available off my home-page.)