[RFC/WIP] unit test for separate debug info

Elena Zannoni ezannoni@redhat.com
Wed Nov 12 14:44:00 GMT 2003


Michael Elizabeth Chastain writes:
 > Hi Elena,
 > 
 > Ah, I was mixing two things together.  I definitely prefer the doco
 > in the format above.  That's one thing.  I was also adding in my own
 > preferences for the file names.  That's a different thing.
 > 

I agree with you about the comment needing to be more explicit.

 > I like gdb.base/break.sym or gdb.base/break.debug a lot more than
 > gdb.base/.debug/break.debug.  That keeps all the files parallel
 > instead of some files inside a dot directory.
 > 

Can't do this one. This is the way that gdb is doing the search, the
default configuration is such that it looks into the .debug
subdirectory.  However, you bring up another valid point, which is
that there is a command in gdb that allows you to change the default
location of the debug files. And I forgot to add that to the test.  I
think I'll do the default location, then move/copy the .debug file and
test the gdb command that changes the location. 

 > 
 > I see your point.  And I see that you see my point.  :)  I guess
 > I'm on your side now.  It's important to test what people actually do.
 > Although if I were building these things as end user of gcc/binutils/gdb,
 > I would build:
 > 
 >   break.full		# full debugging info
 >   break.stripped	# stripped executable
 >   break.sym		# symbols
 >   break.ship		# break.full - symbols + link to break.sym
 >   break			# copy of break.ship
 > 

yes, true. But the distros are coming out with the weird name scheme.
In reality it doesn't make much of a difference. I can change the name
of the final executable in the testcase, and I tried that, no difference.

 > Oh, yeah, I also like the MS-DOS convention of 'break.exe'
 > and I would like to change Unix to do that also.  So I have
 > weird taste.
 > 

er....

elena



More information about the Gdb-patches mailing list