bug#12911: 24.3.50; let users decide where (& perhaps whether)`emacs_backtrace.txt' filesare written

[Drew Adams]

> Again: I understand that you're worried because you've said that
> crashes are almost a daily occurrence for you. But I use the trunk
> [Stefan: not really, I'm using emacs-24, I swear] and I cannot
> remember the last time I had a crash other than assertion failures
> after some sweeping "cleanup".

NOT AT ALL.  Totally irrelevant.  This has nothing to do with me or my setup or
the frequency with which I experience crashes with Emacs 24.

I think you will find no mention by me in the bug report I filed for this, or in
any of my correspondence in the bug thread, of my daily crashes with Emacs 24 or
my concern for my own sake.  Red herring - you are apparently grasping at
straws.

My concern is that Emacs be respectful of users (not particularly me).  For my
personal use I could not care less where Emacs sticks backtrace files.

As to your argument that this is exceptional because Emacs is going down in
flames: I think Stefan has already spoken to that.  We have a choice in that
context regarding how polite Emacs should try to be.

I would say very polite - as polite as Emacs has always been.  You seem to be
saying that we cannot afford to be so polite when Rome is burning.

And I'm guessing you might also say that we cannot even give users the choice as
to how polite Emacs needs to be here, because when you're going down in flames
you cannot be checking user options (we've heard that wrt env vars).

We can agree to disagree about this.  To me, it is more important that Emacs
respect users than it is that Emacs save (and hopefully later send to Emacs Dev)
each and every backtrace.  We've gotten by OK for decades now without such an
annoyance.  And we can apparently (IIUC) find ways to get at least some such
backtrace feedback (i.e, sometimes) without the annoyance.

IIUC, Stefan and Eli have both said that it is only in some cases that Emacs
would not be able to save the backtrace data, if we required Emacs to respect
users and stay out of their folders.  (However, I'm no expert on the
implementation question, and it's quite possible I have misunderstood.)

I say that if that is the case then Emacs Dev should pay the (small) price and
forego those particular backtraces.  Not a big deal, IMHO.  Not as big a deal
as, in effect, telling users that Emacs does not care about their data.