> Probably there's a "dusty deck" somewhere whose inexcusably clever > author relied on the behavior
Well, you can certainly see some JCL somewhere where there is for example // SET DSN1=SOME.PDS(FOO) And there are a bunch of references to DSN=&DSN1, most of which really "want" the member qualification, but one of which is a DISP=(OLD,DELETE) that is in fact working as the coder intended. If I had coded that I don't think I would consider it inexcusably clever. IIRC they did fix another trap for the unwary. A program opens DD1 as an output SAM dataset. Someone executes the program and codes //DD1 DD DSN=SOME.PDS, when what they really meant was SOME.PDS(FOO). IIRC back in the good old days MVS would overwrite the directory of SOME.PDS, with the obvious negative consequences. I think they fixed that and the described scenario now fails. I will leave testing it as an exercise for the reader. > Shell doesn't know what's a file and what's a switch. Well, if it is expanding pax * to pax file file file file could it not recognize that one such file looked a heck of a lot like a switch? Charles -----Original Message----- From: IBM Mainframe Discussion List [mailto:IBM-MAIN@LISTSERV.UA.EDU] On Behalf Of Paul Gilmartin Sent: Thursday, November 5, 2020 12:42 PM To: IBM-MAIN@LISTSERV.UA.EDU Subject: Gwyn's Maxim (was: FSUM7197 pax ...) https://opensource.com/business/14/12/linux-philosophy On Thu, 5 Nov 2020 06:46:15 -0800, Charles Mills wrote: >What the heck were the UNIX designers thinking when they allowed the casual >creation of a filename of -x? There may be a legitimate reason why someone >would want to create a file named -x but if so, then *they* should be made to >jump through some small hoop and "escape" the name in some way. The innocent >victim who stumbles into this situation should not be the one made to jump >through hoops. Will UNIX allow the creation of a file named "rm *"? That could >have some interesting side effects. > Even worse, a file named "-rf *". This resembles the plaint of a JCL novice who has just encountered, painfully, the astonishing behavior of: //SYSUT1 DD DSN=&SYSUID..PDS(MEMBER),DISP=(OLD,DELETE) That should be fixed for DYNALLOC, JCL, TSO, globally by making the TU for member mutex with the TU for delete. Who volunteers to bell the RFE cat? Probably there's a "dusty deck" somewhere whose inexcusably clever author relied on the behavior and whose heirs haven't access to the source. >How did I inadvertently create a file named -x? I had a pax command > >pax -wzvf /my/archive.pax * > >I had an error that I thought might be solved by -x os390. Looking at the >above command I forgot that /my/archive.pax "went with" the -f and coded > >pax -wzvf -x os390 /my/archive.pax * > ... >Heck, if the shell is going to expand the * then it could generate a warning >"hey, did you know that one of your files has a name that looks just like a >switch?" > Shell doesn't know what's a file and what's a switch. That's the responsibility of the utility, even as JCL shouldn't know what PARM means: //STEP EXEC PGM=BPXBATCH,PARM='sudo rm -rf /' -- gil ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN