[Bug symtab/13511] New: 'info variable' and 'info functions' very slow and memory cosuming

[Bug symtab/13511] New: 'info variable' and 'info functions' very slow and memory cosuming

[andre.poenitz at nokia dot com]

http://sourceware.org/bugzilla/show_bug.cgi?id=13511

             Bug #: 13511
           Summary: 'info variable' and 'info functions' very slow and
                    memory cosuming
           Product: gdb
           Version: HEAD
            Status: NEW
          Severity: normal
          Priority: P2
         Component: symtab
        AssignedTo: unassigned@sourceware.org
        ReportedBy: andre.poenitz@nokia.com
    Classification: Unclassified


When starting a "system" libreoffice from ubuntu 11.04, and attaching gdb,
'info variables' takes more then a minute to start producing output. 

When doing the same with a debug build of the a Qt based application (I am use
Qt Creator itself as test case), gdb starts eating all available memory at a
rate of several gigabytes per minute.

-- 
Configure bugmail: http://sourceware.org/bugzilla/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

[Bug symtab/13511] 'info variable' and 'info functions' very slow and memory cosuming

[dje at google dot com]

http://sourceware.org/bugzilla/show_bug.cgi?id=13511

dje at google dot com changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |ccoutant at google dot com,
                   |                            |dje at google dot com

--- Comment #1 from dje at google dot com 2012-01-11 20:25:51 UTC ---
gdb does full symtab expansion here which will use a lot of debug info.

One thought is to have enough information to be able to avoid that.
E.g., could .gdb_index include enough info to avoid having to do full symtab
expansion? [that also has problems, but "info var|fun|type" is common enough
that triggering an explosion of debug info reading is annoying and worth
avoiding]
[and similarly for psymtabs, as well as maybe enhancing pubnames/pubtypes]

Another thought is to be more selective with debug info reading.
Is it possible to read only the parts of the CU (dwarf-speak) that we need?

Another thought is to throw out the info as we're finished using it, perhaps
guided by various parameters.
[we'll have to read it again, but that's preferable to bringing a machine to a
crawl]

-- 
Configure bugmail: http://sourceware.org/bugzilla/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.

[Bug symtab/13511] 'info variable' and 'info functions' very slow and memory cosuming

[tromey at sourceware dot org]

https://sourceware.org/bugzilla/show_bug.cgi?id=13511

Tom Tromey <tromey at sourceware dot org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |tromey at sourceware dot org

--- Comment #2 from Tom Tromey <tromey at sourceware dot org> ---
Maybe this would be more bearable with lazy expansion, see bug #29398

-- 
You are receiving this mail because:
You are on the CC list for the bug.