bug#13775: 24.3.50; Omissions in documentation for crash reporting

bug#13775: 24.3.50; Omissions in documentation for crash reporting

[Dmitry Gutov]

As a non-GDB-wielding user with not much C experience, I had a hard time
following the instructions in `report-emacs-bug' and the ones that
followed.

1. Calling `xbacktrace' requires src/.gdbinit to be loaded. It
a) requires the user to run gdb exactly from src/ (not `gdb src/emacs'),
b) requires them to modify the `auto-load safe-path', or that .gdbinit
is ignored.

2. "Compile without optimizations" - how do I do that? `configure
--help' doesn't seem to show any pertinent options. ...but wait, it says
I can override the choices made by the script.
a) Do I set the variable when calling `make', or do I have to re-run
./configure? Not obvious, the answer is "the latter".
b) I don't know the choice the script made, how do I not break
anything by overriding it? My first choice was "-O0", and that resulted
in an unoptimized build with no debugging symbols (I think).

3. In #13749 (which caused me to write this), Paul also suggests using
-DENABLE_CHECKING. If I'm not mistaken, this variable isn't documented
anywhere.

bug#13775: 24.3.50; Omissions in documentation for crash reporting

[Andreas Schwab]

Dmitry Gutov <dgutov@yandex.ru> writes:

> 3. In #13749 (which caused me to write this), Paul also suggests using
> -DENABLE_CHECKING. If I'm not mistaken, this variable isn't documented
> anywhere.

./configure --enable-checking

Andreas.

-- 
Andreas Schwab, schwab@linux-m68k.org
GPG Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

bug#13775: 24.3.50; Omissions in documentation for crash reporting

[Eli Zaretskii]

> From: Dmitry Gutov <dgutov@yandex.ru>
> Date: Thu, 21 Feb 2013 11:27:06 +0400
> 
> As a non-GDB-wielding user with not much C experience, I had a hard time
> following the instructions in `report-emacs-bug' and the ones that
> followed.

I'm not sure your expectations can be met with any reasonable effort.
There's no way we can teach everybody the basics of GNU-style
configure scripts, including how to discover the available switches
and override compiler and linker options.  You need to learn that
elsewhere.  In general, if you cannot figure out how to do something,
just ask, and you _will_ be helped.

Some answers below, because most of this info is available, you just
need to know how to find it.

> 1. Calling `xbacktrace' requires src/.gdbinit to be loaded. It
> a) requires the user to run gdb exactly from src/ (not `gdb src/emacs'),

The file etc/DEBUG tells you that at the beginning:

  ** When you debug Emacs with GDB, you should start it in the directory
  where the executable was made.  That directory has a .gdbinit file
  that defines various "user-defined" commands for debugging Emacs.
  (These commands are described below under "Examining Lisp object
  values" and "Debugging Emacs Redisplay problems".)

> b) requires them to modify the `auto-load safe-path', or that .gdbinit
> is ignored.

This "feature" entered GDB only recently.  Versions of GDB before 7.5
don't need that, and will barf if you use this command.  I don't see
any reasonable way of dealing with this without confusing newbies even
more (while veteran GDB users already know how to negotiate this
obstacle).

> 2. "Compile without optimizations" - how do I do that? `configure
> --help' doesn't seem to show any pertinent options. ...but wait, it says
> I can override the choices made by the script.

Yes, it says near the end how to specify compiler switches.

> a) Do I set the variable when calling `make', or do I have to re-run
> ./configure? Not obvious, the answer is "the latter".

Actually, both will work.

> b) I don't know the choice the script made, how do I not break
> anything by overriding it?

You just need to know, or ask.  It is not reasonable to expect Emacs
documentation to repeat the entire compiler manual, or large portions
thereof.

> 3. In #13749 (which caused me to write this), Paul also suggests using
> -DENABLE_CHECKING. If I'm not mistaken, this variable isn't documented
> anywhere.

As Andreas points out, this is in "./configure --help".

bug#13775: 24.3.50; Omissions in documentation for crash reporting

[Dmitry Gutov]

On 21.02.2013 21:24, Eli Zaretskii wrote:
> I'm not sure your expectations can be met with any reasonable effort.

If you think that these complaints are not particularly important, feel 
free to close the bug.
Obviously, I've solved all items here for myself already.

> There's no way we can teach everybody the basics of GNU-style
> configure scripts, including how to discover the available switches
> and override compiler and linker options.  You need to learn that\

If by discovering switches you mean calling ./configure --help, then it 
wasn't my problem.
I've missed the --enable-checking switch, but that was later.

The ./configure --help output tells how to override switches in general, 
my complaint is about insufficient detail. Is the "Some influential 
environment variables" part provided by autoconf or somesuch? I haven't 
seen this specific message before.

> elsewhere.  In general, if you cannot figure out how to do something,
> just ask, and you _will_ be helped.

Thanks, I don't doubt that.

> Some answers below, because most of this info is available, you just
> need to know how to find it.
>
>> 1. Calling `xbacktrace' requires src/.gdbinit to be loaded. It
>> a) requires the user to run gdb exactly from src/ (not `gdb src/emacs'),
>
> The file etc/DEBUG tells you that at the beginning:
>
>    ** When you debug Emacs with GDB, you should start it in the directory
>    where the executable was made.  That directory has a .gdbinit file
>    that defines various "user-defined" commands for debugging Emacs.
>    (These commands are described below under "Examining Lisp object
>    values" and "Debugging Emacs Redisplay problems".)

Um, yes, I read that. Maybe I should've skipped this part of the 
complaint. But is this exact wording ("the directory where the 
executable was made") important? If it just said "./src", that would be 
more obvious.

>> b) requires them to modify the `auto-load safe-path', or that .gdbinit
>> is ignored.
>
> This "feature" entered GDB only recently.  Versions of GDB before 7.5
> don't need that, and will barf if you use this command.  I don't see
> any reasonable way of dealing with this without confusing newbies even
> more (while veteran GDB users already know how to negotiate this
> obstacle).

If the feature isn't considered for removal, this argument will become 
less and less important over time. And the odds of a newbie being 
confused by safe-path will approach 100%.

I'm not specifically asking to list the exact commands or ~/.gdbinit 
contents to work around safe-path. Maybe just mention the feature and, 
optionally, suggest consulting GDB manual, if that isn't obvious 
already? But specifying exactly what to do if GDB version is >= 7.5 
would also work.

>> 2. "Compile without optimizations" - how do I do that? `configure
>> --help' doesn't seem to show any pertinent options. ...but wait, it says
>> I can override the choices made by the script.
>
> Yes, it says near the end how to specify compiler switches.
>
>> a) Do I set the variable when calling `make', or do I have to re-run
>> ./configure? Not obvious, the answer is "the latter".
>
> Actually, both will work.

Not exactly.

'CFLAGS="-g3" ./configure' works.
'CFLAGS="-g3" make' doesn't.

'make CFLAGS="-g3"' does work, but AFAIK that's not the usual way of 
binding an environment variable value.

>> b) I don't know the choice the script made, how do I not break
>> anything by overriding it?
>
> You just need to know, or ask.  It is not reasonable to expect Emacs
> documentation to repeat the entire compiler manual, or large portions
> thereof.

I think "compile without optimizations" or "compile for debugging" is a 
sufficiently common special case to warrant listing the recommended 
command somewhere in etc/DEBUG. That will take a few lines at the most.

>> 3. In #13749 (which caused me to write this), Paul also suggests using
>> -DENABLE_CHECKING. If I'm not mistaken, this variable isn't documented
>> anywhere.
>
> As Andreas points out, this is in "./configure --help".

Indeed, thanks for that.

bug#13775: 24.3.50; Omissions in documentation for crash reporting

[Eli Zaretskii]

> Date: Fri, 22 Feb 2013 05:31:55 +0400
> From: Dmitry Gutov <dgutov@yandex.ru>
> CC: 13775@debbugs.gnu.org
> 
> > There's no way we can teach everybody the basics of GNU-style
> > configure scripts, including how to discover the available switches
> > and override compiler and linker options.  You need to learn that\
> 
> If by discovering switches you mean calling ./configure --help, then it 
> wasn't my problem.

I meant both the switches of configure and the compiler/linker
switches and options.

> The ./configure --help output tells how to override switches in general, 
> my complaint is about insufficient detail. Is the "Some influential 
> environment variables" part provided by autoconf or somesuch?

I don't understand the question.  The relevant part of the help text
is this:

  Some influential environment variables:
    CC          C compiler command
    CFLAGS      C compiler flags
    LDFLAGS     linker flags, e.g. -L<lib dir> if you have libraries in a
		nonstandard directory <lib dir>
    LIBS        libraries to pass to the linker, e.g. -l<library>
    CPPFLAGS    (Objective) C/C++ preprocessor flags, e.g. -I<include dir> if
		you have headers in a nonstandard directory <include dir>
    CPP         C preprocessor
    XMKMF       Path to xmkmf, Makefile generator for X Window System

  Use these variables to override the choices made by `configure' or to help
  it to find libraries and programs with nonstandard names/locations.

Which part(s) of this are unclear?

In any case, this is a standard text shown by every configure script
out there, so if you think it should be clarified, please complain to
the Autoconf developers.

> >> 1. Calling `xbacktrace' requires src/.gdbinit to be loaded. It
> >> a) requires the user to run gdb exactly from src/ (not `gdb src/emacs'),
> >
> > The file etc/DEBUG tells you that at the beginning:
> >
> >    ** When you debug Emacs with GDB, you should start it in the directory
> >    where the executable was made.  That directory has a .gdbinit file
> >    that defines various "user-defined" commands for debugging Emacs.
> >    (These commands are described below under "Examining Lisp object
> >    values" and "Debugging Emacs Redisplay problems".)
> 
> Um, yes, I read that. Maybe I should've skipped this part of the 
> complaint. But is this exact wording ("the directory where the 
> executable was made") important? If it just said "./src", that would be 
> more obvious.

I added that (revision 111290 on the emacs-24 branch).

> >> b) requires them to modify the `auto-load safe-path', or that .gdbinit
> >> is ignored.
> >
> > This "feature" entered GDB only recently.  Versions of GDB before 7.5
> > don't need that, and will barf if you use this command.  I don't see
> > any reasonable way of dealing with this without confusing newbies even
> > more (while veteran GDB users already know how to negotiate this
> > obstacle).
> 
> If the feature isn't considered for removal, this argument will become 
> less and less important over time. And the odds of a newbie being 
> confused by safe-path will approach 100%.

But GDB already tells you how to allow .gdbinit to be auto-loaded, and
also points to the GDB manual.  If the text displayed by GDB is not
clear or confusing, I suggest to report that to GDB maintainers.

> I'm not specifically asking to list the exact commands or ~/.gdbinit 
> contents to work around safe-path. Maybe just mention the feature and, 
> optionally, suggest consulting GDB manual, if that isn't obvious 
> already?

I added that to etc/DEBUG.

> But specifying exactly what to do if GDB version is >= 7.5 
> would also work.

That's hard to do, because the solution depends on the end-user's
preferences regarding security and on the degree of their machine's
exposure to other users and to the outside world.  The GDB manual
discusses the possible solutions, so a pointer to it will allow the
user to make up her mind.

> >> a) Do I set the variable when calling `make', or do I have to re-run
> >> ./configure? Not obvious, the answer is "the latter".
> >
> > Actually, both will work.
> 
> Not exactly.
> 
> 'CFLAGS="-g3" ./configure' works.
> 'CFLAGS="-g3" make' doesn't.
> 
> 'make CFLAGS="-g3"' does work, but AFAIK that's not the usual way of 
> binding an environment variable value.

CFLAGS is a Make variable.  Make normally initializes all its
variables by looking at the environment.  But 'CFLAGS="-g3" make'
doesn't export the value of CFLAGS for Make to see it, it only inserts
CFLAGS into the shell's own environment.  That is why the command
'CFLAGS="-g3" make' doesn't work, while 'make CFLAGS="-g3"' does.

This is all standard shell and Make stuff, I don't think it's
reasonable to expect Emacs documentation to teach all that.

> I think "compile without optimizations" or "compile for debugging" is a 
> sufficiently common special case to warrant listing the recommended 
> command somewhere in etc/DEBUG. That will take a few lines at the most.

It's already there, it just didn't mention the -O0 option explicitly;
I added that.  (Again, this is a basic compiler option, not something
specific to Emacs.)

bug#13775: 24.3.50; Omissions in documentation for crash reporting

[Andreas Schwab]

Dmitry Gutov <dgutov@yandex.ru> writes:

> 'CFLAGS="-g3" make' doesn't.

As expected and documented.

> 'make CFLAGS="-g3"' does work, but AFAIK that's not the usual way of
> binding an environment variable value.

This is wrong.  Environment variables never override makefile variables.

Andreas.

-- 
Andreas Schwab, schwab@linux-m68k.org
GPG Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

bug#13775: 24.3.50; Omissions in documentation for crash reporting

[Dmitry Gutov]

On 22.02.2013 13:37, Andreas Schwab wrote:
> Dmitry Gutov <dgutov@yandex.ru> writes:
>
>> 'CFLAGS="-g3" make' doesn't.
>
> As expected and documented.
>
>> 'make CFLAGS="-g3"' does work, but AFAIK that's not the usual way of
>> binding an environment variable value.
>
> This is wrong.  Environment variables never override makefile variables.

Sorry, what's wrong, exactly? My expectations, my retelling, or the way 
it actually works?

bug#13775: 24.3.50; Omissions in documentation for crash reporting

[Eli Zaretskii]

> Date: Sat, 23 Feb 2013 01:26:21 +0400
> From: Dmitry Gutov <dgutov@yandex.ru>
> CC: Eli Zaretskii <eliz@gnu.org>, 13775@debbugs.gnu.org
> 
> On 22.02.2013 13:37, Andreas Schwab wrote:
> > Dmitry Gutov <dgutov@yandex.ru> writes:
> >
> >> 'CFLAGS="-g3" make' doesn't.
> >
> > As expected and documented.
> >
> >> 'make CFLAGS="-g3"' does work, but AFAIK that's not the usual way of
> >> binding an environment variable value.
> >
> > This is wrong.  Environment variables never override makefile variables.
> 
> Sorry, what's wrong, exactly? My expectations, my retelling, or the way 
> it actually works?

Your expectations that putting a variable into the environment will
override Make variable that is assigned in the Makefile.  (If a
variable is not assigned in the Makefile, but used by one of its
rules, Make will use the value in the environment.)

And anyway, 'CFLAGS="-g3" make' doesn't export CFLAGS into the
environment of programs run by the shell, so this method is simply
wrong, for reasons that have nothing to do with Emacs.

bug#13775: 24.3.50; Omissions in documentation for crash reporting

[Dmitry Gutov]

On 22.02.2013 13:24, Eli Zaretskii wrote:
> I meant both the switches of configure and the compiler/linker
> switches and options.
>
>> The ./configure --help output tells how to override switches in general,
>> my complaint is about insufficient detail. Is the "Some influential
>> environment variables" part provided by autoconf or somesuch?
>
> I don't understand the question.  The relevant part of the help text

You answered it below, "this is a standard text shown by ...".

> is this:
>
>    Some influential environment variables:
>      CC          C compiler command
>      CFLAGS      C compiler flags
>      LDFLAGS     linker flags, e.g. -L<lib dir> if you have libraries in a
> 		nonstandard directory <lib dir>
>      LIBS        libraries to pass to the linker, e.g. -l<library>
>      CPPFLAGS    (Objective) C/C++ preprocessor flags, e.g. -I<include dir> if
> 		you have headers in a nonstandard directory <include dir>
>      CPP         C preprocessor
>      XMKMF       Path to xmkmf, Makefile generator for X Window System
>
>    Use these variables to override the choices made by `configure' or to help
>    it to find libraries and programs with nonstandard names/locations.
>
> Which part(s) of this are unclear?
>
> In any case, this is a standard text shown by every configure script
> out there, so if you think it should be clarified, please complain to
> the Autoconf developers.

Thanks, maybe I'll do that, sometime.

>>>> 1. Calling `xbacktrace' requires src/.gdbinit to be loaded. It
>>>> a) requires the user to run gdb exactly from src/ (not `gdb src/emacs'),
>>>
>>> The file etc/DEBUG tells you that at the beginning:
>>>
>>>     ** When you debug Emacs with GDB, you should start it in the directory
>>>     where the executable was made.  That directory has a .gdbinit file
>>>     that defines various "user-defined" commands for debugging Emacs.
>>>     (These commands are described below under "Examining Lisp object
>>>     values" and "Debugging Emacs Redisplay problems".)
>>
>> Um, yes, I read that. Maybe I should've skipped this part of the
>> complaint. But is this exact wording ("the directory where the
>> executable was made") important? If it just said "./src", that would be
>> more obvious.
>
> I added that (revision 111290 on the emacs-24 branch).
>
>>>> b) requires them to modify the `auto-load safe-path', or that .gdbinit
>>>> is ignored.
>>>
>>> This "feature" entered GDB only recently.  Versions of GDB before 7.5
>>> don't need that, and will barf if you use this command.  I don't see
>>> any reasonable way of dealing with this without confusing newbies even
>>> more (while veteran GDB users already know how to negotiate this
>>> obstacle).
>>
>> If the feature isn't considered for removal, this argument will become
>> less and less important over time. And the odds of a newbie being
>> confused by safe-path will approach 100%.
>
> But GDB already tells you how to allow .gdbinit to be auto-loaded, and
> also points to the GDB manual.  If the text displayed by GDB is not
> clear or confusing, I suggest to report that to GDB maintainers.
>
>> I'm not specifically asking to list the exact commands or ~/.gdbinit
>> contents to work around safe-path. Maybe just mention the feature and,
>> optionally, suggest consulting GDB manual, if that isn't obvious
>> already?
>
> I added that to etc/DEBUG.
>
>> But specifying exactly what to do if GDB version is >= 7.5
>> would also work.
>
> That's hard to do, because the solution depends on the end-user's
> preferences regarding security and on the degree of their machine's
> exposure to other users and to the outside world.  The GDB manual
> discusses the possible solutions, so a pointer to it will allow the
> user to make up her mind.
>
>>>> a) Do I set the variable when calling `make', or do I have to re-run
>>>> ./configure? Not obvious, the answer is "the latter".
>>>
>>> Actually, both will work.
>>
>> Not exactly.
>>
>> 'CFLAGS="-g3" ./configure' works.
>> 'CFLAGS="-g3" make' doesn't.
>>
>> 'make CFLAGS="-g3"' does work, but AFAIK that's not the usual way of
>> binding an environment variable value.
>
> CFLAGS is a Make variable.  Make normally initializes all its
> variables by looking at the environment.  But 'CFLAGS="-g3" make'
> doesn't export the value of CFLAGS for Make to see it, it only inserts
> CFLAGS into the shell's own environment.  That is why the command
> 'CFLAGS="-g3" make' doesn't work, while 'make CFLAGS="-g3"' does.

I see, I think.

> This is all standard shell and Make stuff, I don't think it's
> reasonable to expect Emacs documentation to teach all that.
>
>> I think "compile without optimizations" or "compile for debugging" is a
>> sufficiently common special case to warrant listing the recommended
>> command somewhere in etc/DEBUG. That will take a few lines at the most.
>
> It's already there, it just didn't mention the -O0 option explicitly;
> I added that.  (Again, this is a basic compiler option, not something
> specific to Emacs.)

Like I described, my difficulty here was that using "-O0" by itself is 
not enough, without "-g3" the resulting build is even less debuggable.

Omitting "-O0", leaving just "-g3", works, on the other hand, because 
the optimization is enabled by default due to the default CFLAGS value 
(AFAICT).

Anyway, you've addressed that in 111290, as well as two other items. 
Thank you, I'm considering this issue fixed.

bug#13775: 24.3.50; Omissions in documentation for crash reporting

[Dmitry Gutov]

On 23.02.2013 1:44, Eli Zaretskii wrote:
>> Date: Sat, 23 Feb 2013 01:26:21 +0400
>> From: Dmitry Gutov <dgutov@yandex.ru>
>> CC: Eli Zaretskii <eliz@gnu.org>, 13775@debbugs.gnu.org
>>
>> On 22.02.2013 13:37, Andreas Schwab wrote:
>>> Dmitry Gutov <dgutov@yandex.ru> writes:
>>>
>>>> 'CFLAGS="-g3" make' doesn't.
>>>
>>> As expected and documented.
>>>
>>>> 'make CFLAGS="-g3"' does work, but AFAIK that's not the usual way of
>>>> binding an environment variable value.
>>>
>>> This is wrong.  Environment variables never override makefile variables.
>>
>> Sorry, what's wrong, exactly? My expectations, my retelling, or the way
>> it actually works?
>
> Your expectations that putting a variable into the environment will
> override Make variable that is assigned in the Makefile.  (If a
> variable is not assigned in the Makefile, but used by one of its
> rules, Make will use the value in the environment.)

I expected it may get picked up in some special way. ./configure does 
that, after all, and it also shells out during execution.

> And anyway, 'CFLAGS="-g3" make' doesn't export CFLAGS into the
> environment of programs run by the shell, so this method is simply
> wrong, for reasons that have nothing to do with Emacs.

So, I was doubly wrong. Thanks for elaboration. :)