Ran across that at a client. I believe that the PL/I manuals document that an LE enabled program cannot call a PL/I MAIN.
On Tue, 27 Mar 2012 18:16:06 +0200 Bernd Oppolzer <[email protected]> wrote: :>Hello all, :> :>we have a problem that is not easy to describe. Let me try it. :> :>We are building a test supporting system, which allows to do tests of :>software :>components. The system consists of the following modules or parts: :> :>A - driver, written in ASSEMBLER. The driver builds a LE enclave, so :>that it :>can call C subfunctions and the test objects, which are (normally) PL/1 :>routines :> :>B - driver supporting routine, written in C. This routine provides :>services OPEN, :>WRITE and CLOSE. It is called by the driver. On OPEN, it fetches a :>testcase list :>from the testcase database and builds a linked list in memory (in the LE :>heap). :>It also opens a log file for writing. On WRITE, it writes some protocol data :>on the log file. On CLOSE, it closes the log file. :> :>C - driver exit, written in ASSEMBLER. It is called at the beginning of :>the test object :>(at the entry point) by means of TRAP2 interrupts. It has access to the :>testcase list :>element that B took from the testcase database for the current testcase. :>It then calls :>D, passing the address of the testcase list element. D reads the :>testcase data, :>builds a parameter list for the testcase, and returns a parameter list. :>Then the :>driver exits resumes the test object at the position of the entry point :>(this way :>the test data is passed to the test object). :> :>D - is a C routine which reads the data from the testcase data base and :>builds the :>parameter list for the test object - because this is much easier to do :>in C than in :>ASSEMBLER. :> :>Now the problem: :> :>all works well, if the test object is a PL/1 subroutine and not a main :>program. Then :>the LE enclave built by A is used by B, passed to the test object, and :>used by D as well. :>There is no problem passing the LE environment through the TRAP2 interrupt. :> :>But: :> :>we also want to be able to test PL/1 main programs this way. The PL/1 :>main programs :>are called by the driver the same way as the modules are called, but :>because they are :>started at CEESTART, a second LE enclave below the first enclave is :>built. Then C and :>D run below the second enclave. This is no problem so far, because there :>is no need :>to share ressources between the two enclaves; the only information :>exchange is through :>the testcase list element which belongs to the heap of the first :>enclave, and this is no problem, :>in our opinion. :> :>But: :> :>when we return from the test object and we try to execute the WRITE call :>to B in the :>first enclave, it fails in the prologue of B. :> :>We first thought that reg 12 had not be restored correctly, but this is :>not the case. :>Reg 12 is the same as before the call of the PL/1 main, points to the :>original CEECAA. :>Also, the save area (including NAB etc, the words behind offsets 72) are :>unchanged. :>But still, we get abends 0C4 etc. in the prologue of B. But only in the :>PL/1 main case, :>if there is a second enclave, not in the other case. :> :>We fixed this problem by doing the WRITE and CLOSE calls not in the same :>(first) :>enclave as the OPEN call, but instead we did all the subsequent calls by :>an intermediate :>PL/1 main, that is: after the PL/1 test object main destroyed somehow :>our first enclave, :>we didn't use it any more, but instead constructed new enclaves for :>every subsequent :>call at the same level. But this doesn't look very sound to me. :> :>My question is: has somebody out there an idea, what has happened here :>and what :>might be a solution to our problem? Is it possible to have one LE :>enclave call another :>and on return from the second still have the first one usable? What are :>we getting wrong? :> :>Kind regards :> :>Bernd :> :>---------------------------------------------------------------------- :>For IBM-MAIN subscribe / signoff / archive access instructions, :>send email to [email protected] with the message: INFO IBM-MAIN -- Binyamin Dissen <[email protected]> http://www.dissensoftware.com Director, Dissen Software, Bar & Grill - Israel Should you use the mailblocks package and expect a response from me, you should preauthorize the dissensoftware.com domain. I very rarely bother responding to challenge/response systems, especially those from irresponsible companies. ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: INFO IBM-MAIN
