Bug 6632

Summary:	Debugging fhpd with fhpd does not print source
Product:	frysk	Reporter:	Sami Wagiaalla <swagiaal>
Component:	general	Assignee:	Unassigned <frysk-bugzilla>
Status:	NEW ---
Severity:	normal	CC:	swagiaal
Priority:	P2
Version:	unspecified
Target Milestone:	---
Host:		Target:
Build:		Last reconfirmed:
Project(s) to access:		ssh public key:
Bug Depends on:
Bug Blocks:	2246

Description Sami Wagiaalla 2008-06-11 19:48:08 UTC

[swagiaal@toner frysk.bugs]$ fhpd ./frysk-core/frysk/bindir/fhpd $$
[0.0] Loaded executable file:
/notnfs/swagiaal/frysks.bin/frysk.bugs/frysk-core/frysk/bindir/fhpd
(fhpd) break main
breakpoint 0 deferred
(fhpd) run
running with this command:
/notnfs/swagiaal/frysks.bin/frysk.bugs/frysk-core/frysk/bindir/fhpd 3669
Attached to process 4335
Running process 4335
(fhpd) Breakpoint 0 main 0x4c2b70
list
[0.0]
Error: file /tmp/ccD9MbpQ.i not found.
(fhpd)

Comment 1 Andrew Cagney 2008-06-11 20:40:20 UTC

What's happening is that GCJ generates a main() stub which will then call the
java runtime starting up FHPD.  The generated main() lives in a temp file such
as the below.

The DWARF committee keep considering ways to address this; for instance how to
describe, in dwarf, something that really isn't debug-info but rather run-time info.

Anyway, if we could find the true entry point, then "run" could automatically
breakpoint in "main".

btw #1; fortran and ada have the same problem
btw #2: .stabs (another debug format) were, at one stage genrating an N_main stab.

perhaps check the DWARF web site to see if anything got resolved.