Re: [Libreoffice] Even more debugging info

On Thursday 01 of December 2011, Lionel Elie Mamane wrote:
> On Thu, Dec 01, 2011 at 12:14:32PM +0100, Lubos Lunak wrote:
> > On Wednesday 30 of November 2011, Tom Tromey wrote:
> > > First, one must consider the tradeoffs.  I always use -g3 when building
> > > gdb, because gdb uses macros fairly heavily, and because I just don't
> > > care about the extra space.  I prefer the convenience.  YMMV of course,
> > > and I don't know enough about LibreOffice to offer an opinion.
> >
> >  There is hope that LibreOffice is C++ enough to actually not use
> > macros that need to be stepped through.
>
> Well, I find it useful that when I'm on a line like
>
>    foo ( bar, SOME_MACRO )
>
> I can do "print SOME_MACRO" and get the value without hunting down the
> #define SOME_MACRO 1022
>
> And LibreOffice definitely uses a lot of such #defines, and not
>  const int SOME_VALUE = 1022;

 Well, I said that's what the hope would be, I wouldn't be as naive as to 
think that would be the exact reality. Still, on the whole, a better solution 
to your problem is probably submitting a bunch of s/#define/const int/ 
patches where needed rather than making the build even larger and slower for 
everybody. At least, you do not have any numbers to prove otherwise.

On Thursday 01 of December 2011, Lionel Elie Mamane wrote:
> On Wed, Nov 30, 2011 at 11:31:19PM +0100, Michael Stahl wrote:
> > an interesting data point:
> > LO tree built with gcc 4.6.2 -g takes up 23G.
> >
> > extrapolating from Lubos' mail that -ggdb3 takes 4 times as much space
> > the problem should be obvious :-/
>
> I'm not sure what/how you are measuring,

 Apparently, the size of the whole LO tree, after running a build there.

> but with -gdb3: 
>
> $ du --apparent-size -csh workdir/ solver/
> 22G   workdir/
> 21G   solver/

 This actually proves the point that -ggdb3 significantly increases the build 
size (and thus also build time). You do not provide any values to compare 
these against, so they are not very conclusive, but when compared to 
Michael's number, your LO tree is double in size, at the very least. Newer 
gcc versions and -gdwarf-4 may provide more compact debug info, so it may not 
be 4x as in my case from the half a year ago, but IMO that's still a lot, for 
the default.

-- 
 Lubos Lunak
 l.lunak@suse.cz