home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!sun-barr!cs.utexas.edu!ut-emx!ibmchs!auschs!awdprime.austin.ibm.com!yogi.austin.ibm.com!marc
- From: marc@yogi.austin.ibm.com (Marc J. Stephenson)
- Newsgroups: comp.unix.aix
- Subject: Re: dbx and 'load'ed code
- Message-ID: <1992Sep8.144318.22105@awdprime.austin.ibm.com>
- Date: 8 Sep 92 14:43:18 GMT
- References: <MedW01O00WoiEsTkJ1@andrew.cmu.edu>
- Sender: news@awdprime.austin.ibm.com (USENET News)
- Organization: IBM, Austin
- Lines: 28
-
- In article <MedW01O00WoiEsTkJ1@andrew.cmu.edu> rr2b+@andrew.cmu.edu (Robert Andrew Ryan) writes:
- >Is there any way to make dbx look somewhere other than the current
- >directory for object files which the program loads with load? The
- >objects are loaded with an absolute path name, and no library path.
- >Whenever I run the program under dbx dbx dies unless the loaded objects
- >are in the current directory. Pointers to the relevant part of TFM
- >welcome, I looked...
- >
- >Thanks for any info,
- >-Rob
-
- You've either run into a dbx bug or a loader bug. There is no special
- option to tell dbx where to pick up loaded code. When dbx detects a dynamic
- load, it asks the loader (through ptrace) where all the pieces of the program
- are (some pieces could have been unloaded as well). The loader gives back
- the information in two forms - the name of the program (along with a library
- name if it is part of an archive) and a valid open file descriptor. If opening
- the object file (to get symbolic information) by the default way fails (I think
- that it tries to open by name first, not that it really matters), then it tries
- to open the object by the other means. If that fails, then dbx is sunk in
- trying to open the object file (barring invention of some other means).
-
- So, either the logic in dbx is toast, or the loader is sending back wrong
- information to dbx. Call in a defect through your customer support.
- --
- Marc Stephenson IBM AWD (Advanced Workstations Division - Austin,TX)
- DISCLAIMER: The content of this posting is independent of official IBM position.
- INTERNET->marc@perfmap.austin.ibm.com VNET: MARC at AUSVMQ IBM T/L: 678-3189
-