home *** CD-ROM | disk | FTP | other *** search
- Comments: Gated by NETNEWS@AUVM.AMERICAN.EDU
- Path: sparky!uunet!stanford.edu!bcm!convex!darwin.sura.net!paladin.american.edu!auvm!WEIZMANN.WEIZMANN.AC.IL!VSLAWR
- Message-ID: <MAILBOOK%92111009135883@DEARN>
- Newsgroups: bit.listserv.mailbook
- Date: Tue, 10 Nov 1992 09:39:33 +0200
- Sender: RiceMail discussion list <MAILBOOK@RICEVM1.BITNET>
- From: "Lawrence S. Kalman" <VSLAWR@WEIZMANN.WEIZMANN.AC.IL>
- Subject: Re: Compiling MAIL
- In-Reply-To: Message of Mon,
- 9 Nov 92 10:59:20 CST from <SCHAFER@RICEVM1.RICE.EDU>
- Lines: 20
-
- On Mon, 9 Nov 92 10:59:20 CST Richard A. Schafer said:
- > Sourceline will make the file bigger, but will also allow
- >your users to give you better feedback should something go wrong in the code;
- >the choice is yours to make which is worth more to you.
-
- We don't compile our MAIL/MAILBOOK code, but we generate a "preparsed"
- version (using a technique which uses the Rexx Interpreter itself to do the
- generation) which is loaded into CMSINST. We get a measuarable performance
- improvement (6-8%) at the expense of rendering the "source" code totally
- unintelligble for debugging purposes. Our solution is to maintain a
- non-preparsed and uncompressed (one statement per line) version of the execs
- on a disk accessed at the mode specified in SET INSTSEG ON (in our case, P).
- In normal operation, the user is running the preparsed and compressed version
- from CMSINST. If a problem occurs, the user is told to issue SET INSTSEG OFF
- and run MAIL/MAILBOOK again. The resulting traceback is much more informative
- than that obtained in the normal operation of the compressed exec, because the
- statement in error is precisely identified, rather than the traceback pointing
- to a line containing numerous statements.
-
- - Lawrence
-