Re: Debugging emacs memory management

[Eli Zaretskii]

> From: Dima Kogan <address@hidden>
> Cc: Eli Zaretskii <address@hidden>, address@hidden, address@hidden
> Date: Mon, 05 Oct 2015 03:02:16 -0700
> 
> Andreas Schwab <address@hidden> writes:
> > Dima Kogan <address@hidden> writes:
> >
> >> diff --git a/src/font.c b/src/font.c
> >> index 8e06532..ca872d0 100644
> >> --- a/src/font.c
> >> +++ b/src/font.c
> >> @@ -3981,7 +3981,15 @@ copy_font_spec (Lisp_Object font)
> >>    pcdr = spec->props + FONT_EXTRA_INDEX;
> >>    for (tail = AREF (font, FONT_EXTRA_INDEX); CONSP (tail); tail = XCDR 
> >> (tail))
> >>      if (!EQ (XCAR (XCAR (tail)), QCfont_entity))
> >> -      *pcdr = Fcons (XCAR (tail), Qnil), pcdr = xcdr_addr (*pcdr);
> >> +      {
> >> +        if (CONSP (XCAR (tail)))
> >
> > This better be always true, otherwise XCAR (XCAR (tail)) would be a bug.
> 
> It isn't. Without that check I was getting every time.
> 
>   *ERROR*: Wrong type argument: listp, "monospace-10"

I think what Andreas meant was this: the old code used
XCAR (XCAR (tail)) unconditionally, which seems to indicate that
XCAR (tail) is always a cons cell, and your test should always
yield true.

> Surprised me too, and I haven't checked to see what was happening there.

So maybe your code should include some check for QCfont_entity in the
'else' clause as well.

Btw, isn't 12KB per frame still a lot?

Thanks.