gnutls error

gnutls error

[Colin Baxter]

Hello,

I have the following error at gnutls.o when I attempt to make freshly
pulled emacs-26. 

---------- Begin ---

gnutls.c: In function ‘syms_of_gnutls’:
gnutls.c:2521:13: error: ‘Sgnutls_ciphers’ undeclared (first use in this function)
gnutls.c:2521:13: note: each undeclared identifier is reported only once for each function it appears in
gnutls.c:2522:13: error: ‘Sgnutls_macs’ undeclared (first use in this function)
gnutls.c:2523:13: error: ‘Sgnutls_digests’ undeclared (first use in this function)
gnutls.c:2524:13: error: ‘Sgnutls_hash_mac’ undeclared (first use in this function)
gnutls.c:2525:13: error: ‘Sgnutls_hash_digest’ undeclared (first use in this function)
gnutls.c:2526:13: error: ‘Sgnutls_symmetric_encrypt’ undeclared (first use in this function)
gnutls.c:2527:13: error: ‘Sgnutls_symmetric_decrypt’ undeclared (first use in this function)
make[1]: *** [gnutls.o] Error 1

---------- End -----

I have tried reverting recent commits that I thought relevant - but to
no avail.

I use Linux 3.2.0-4-686-pae #1 SMP Debian 3.2.89-2 i686 GNU/Linux.

Best wishes,

Colin.

-- 
--
Colin Baxter
m43cap@yandex.com
GnuPG fingerprint: 68A8 799C 0230 16E7 BF68  2A27 BBFA 2492 91F5 41C8

Re: gnutls error

[Colin Baxter]

I forgot to mention I use gnutls-cli (GnuTLS) 2.12.20
Packaged by Debian (2.12.20-8+deb7u5).



-- 
--
Colin Baxter
m43cap@yandex.com
GnuPG fingerprint: 68A8 799C 0230 16E7 BF68  2A27 BBFA 2492 91F5 41C8

Re: gnutls error

[Eli Zaretskii]

> From: Colin Baxter <m43cap@yandex.com>
> Cc: 
> Date: Sat, 15 Jul 2017 10:49:21 +0100
> 
> I have the following error at gnutls.o when I attempt to make freshly
> pulled emacs-26. 
> 
> ---------- Begin ---
> 
> gnutls.c: In function ‘syms_of_gnutls’:
> gnutls.c:2521:13: error: ‘Sgnutls_ciphers’ undeclared (first use in this function)
> gnutls.c:2521:13: note: each undeclared identifier is reported only once for each function it appears in
> gnutls.c:2522:13: error: ‘Sgnutls_macs’ undeclared (first use in this function)
> gnutls.c:2523:13: error: ‘Sgnutls_digests’ undeclared (first use in this function)
> gnutls.c:2524:13: error: ‘Sgnutls_hash_mac’ undeclared (first use in this function)
> gnutls.c:2525:13: error: ‘Sgnutls_hash_digest’ undeclared (first use in this function)
> gnutls.c:2526:13: error: ‘Sgnutls_symmetric_encrypt’ undeclared (first use in this function)
> gnutls.c:2527:13: error: ‘Sgnutls_symmetric_decrypt’ undeclared (first use in this function)
> make[1]: *** [gnutls.o] Error 1
> 
> ---------- End -----

Please try again, I think I fixed this.

Re: gnutls error

[Dani Moncayo]

I've tried to build the current version of the master branch on
Ubuntu-GNOME 17.04; it fails this way:

[...]
  CC       floatfns.o
  CC       fns.o
In file included from ../../repo/src/fns.c:30:0:
../../repo/src/fns.c:4748:34: error: ‘Fsecure_hash_algorithms’
undeclared here (not in a function)
 DEFUN ("secure-hash-algorithms", Fsecure_hash_algorithms,
                                  ^
../../repo/src/lisp.h:2925:26: note: in definition of macro ‘DEFUN’
        { .a ## maxargs = fnname },     \
                          ^~~~~~
../../repo/src/fns.c:4748:34: warning: no previous prototype for
‘Fsecure_hash_algorithms’ [-Wmissing-prototypes]
 DEFUN ("secure-hash-algorithms", Fsecure_hash_algorithms,
                                  ^
../../repo/src/lisp.h:2927:16: note: in definition of macro ‘DEFUN’
    Lisp_Object fnname
                ^~~~~~
../../repo/src/fns.c: In function ‘extract_data_from_object’:
../../repo/src/fns.c:4928:24: error: ‘Qiv_auto’ undeclared (first use
in this function)
   else if (EQ (object, Qiv_auto))
                        ^~~~~~~~
../../repo/src/fns.c:4928:24: note: each undeclared identifier is
reported only once for each function it appears in
Makefile:374: recipe for target 'fns.o' failed
make[1]: *** [fns.o] Error 1
make[1]: *** Waiting for unfinished jobs....
make[1]: Leaving directory '/home/dani/devel/emacs/build/src'
Makefile:416: recipe for target 'src' failed
make: *** [src] Error 2



-- 
Dani Moncayo

Re: gnutls error

[Eli Zaretskii]

> From: Dani Moncayo <dmoncayo@gmail.com>
> Date: Sat, 15 Jul 2017 13:14:35 +0200
> 
> I've tried to build the current version of the master branch on
> Ubuntu-GNOME 17.04; it fails this way:
> 
> [...]
>   CC       floatfns.o
>   CC       fns.o
> In file included from ../../repo/src/fns.c:30:0:
> ../../repo/src/fns.c:4748:34: error: ‘Fsecure_hash_algorithms’
> undeclared here (not in a function)
>  DEFUN ("secure-hash-algorithms", Fsecure_hash_algorithms,
>                                   ^
> ../../repo/src/lisp.h:2925:26: note: in definition of macro ‘DEFUN’
>         { .a ## maxargs = fnname },     \
>                           ^~~~~~
> ../../repo/src/fns.c:4748:34: warning: no previous prototype for
> ‘Fsecure_hash_algorithms’ [-Wmissing-prototypes]
>  DEFUN ("secure-hash-algorithms", Fsecure_hash_algorithms,
>                                   ^
> ../../repo/src/lisp.h:2927:16: note: in definition of macro ‘DEFUN’
>     Lisp_Object fnname
>                 ^~~~~~
> ../../repo/src/fns.c: In function ‘extract_data_from_object’:
> ../../repo/src/fns.c:4928:24: error: ‘Qiv_auto’ undeclared (first use
> in this function)
>    else if (EQ (object, Qiv_auto))
>                         ^~~~~~~~
> ../../repo/src/fns.c:4928:24: note: each undeclared identifier is

These errors make no sense, the stuff they complain about is defined
right there in fns.c.

How did you build Emacs?  With what sequence of commands?  Are you
sure your Git clone is up to date with upsteam?

FWIW, I've built the current master on both Windows and GNU/Linux, and
didn't see any such errors.

Re: gnutls error

[Colin Baxter]

>>>>> "Eli" == Eli Zaretskii <eliz@gnu.org> writes:

    Eli> Please try again, I think I fixed this.

Yes, the error no longer occurs. Emacs-26 now builds satisfactorily, for
me. Thank you.

Best wishes,

Colin.

-- 
--
Colin Baxter
m43cap@yandex.com
GnuPG fingerprint: 68A8 799C 0230 16E7 BF68  2A27 BBFA 2492 91F5 41C8

Re: gnutls error

[Masaru Nomiya]

Hello,

In the Message; 

  Subject    : Re: gnutls error
  Message-ID : <87shhxex1d.fsf@yandex.com>
  Date & Time: Sat, 15 Jul 2017 12:53:34 +0100

Colin Baxter <m43cap@yandex.com> has written:

> >>>>> "Eli" == Eli Zaretskii <eliz@gnu.org> writes:

>     Eli> Please try again, I think I fixed this.

> Yes, the error no longer occurs. Emacs-26 now builds satisfactorily, for
> me. Thank you.

Did you?

I still got errors with the very latest git head;

[...]
  CCLD     temacs
gnutls.o: In function `Fgnutls_ciphers':
/tmp/tes/emacs/src/gnutls.c:1860: undefined reference to `gnutls_cipher_get_tag_size'
/tmp/tes/emacs/src/gnutls.c:1876: undefined reference to `gnutls_cipher_get_iv_size'
gnutls.o: In function `Fgnutls_macs':
/tmp/tes/emacs/src/gnutls.c:2197: undefined reference to `gnutls_mac_get_nonce_size'
gnutls.o: In function `Fgnutls_digests':
/tmp/tes/emacs/src/gnutls.c:2213: undefined reference to `gnutls_digest_list'
/tmp/tes/emacs/src/gnutls.c:2218: undefined reference to `gnutls_digest_get_name'
gnutls.o: In function `gnutls_symmetric':
/tmp/tes/emacs/src/gnutls.c:2038: undefined reference to `gnutls_cipher_get_iv_size'
/tmp/tes/emacs/src/gnutls.c:2055: undefined reference to `gnutls_cipher_get_tag_size'
collect2: error: ld returned 1 exit status
Makefile:598: recipe for target 'temacs' failed
make[2]: *** [temacs] Error 1
make[2]: Leaving directory '/tmp/tes/emacs/src'
Makefile:416: recipe for target 'src' failed
make[1]: *** [src] Error 2
make[1]: Leaving directory '/tmp/tes/emacs'
Makefile:1097: recipe for target 'bootstrap' failed
make: *** [bootstrap] Error 2

---
 Masaru Nomiya

Re: gnutls error

[Colin Baxter]

Dear Masaru,

>>>>> "Masaru" == Masaru Nomiya <nomiya@galaxy.dti.ne.jp> writes:

    Masaru> Hello, In the Message;

    Masaru>   Subject : Re: gnutls error Message-ID :
    Masaru> <87shhxex1d.fsf@yandex.com> Date & Time: Sat, 15 Jul 2017
    Masaru> 12:53:34 +0100

    Masaru> Colin Baxter <m43cap@yandex.com> has written:

    >> >>>>> "Eli" == Eli Zaretskii <eliz@gnu.org> writes:

    Eli> Please try again, I think I fixed this.

    >> Yes, the error no longer occurs. Emacs-26 now builds
    >> satisfactorily, for me. Thank you.

    Masaru> Did you?

    Masaru> I still got errors with the very latest git head;

    Masaru> [...]  CCLD temacs gnutls.o: In function `Fgnutls_ciphers':
    Masaru> /tmp/tes/emacs/src/gnutls.c:1860: undefined reference to
    Masaru> `gnutls_cipher_get_tag_size'
    Masaru> /tmp/tes/emacs/src/gnutls.c:1876: undefined reference to
    Masaru> `gnutls_cipher_get_iv_size' gnutls.o: In function
    Masaru> `Fgnutls_macs': /tmp/tes/emacs/src/gnutls.c:2197: undefined
    Masaru> reference to `gnutls_mac_get_nonce_size' gnutls.o: In
    Masaru> function `Fgnutls_digests':
    Masaru> /tmp/tes/emacs/src/gnutls.c:2213: undefined reference to
    Masaru> `gnutls_digest_list' /tmp/tes/emacs/src/gnutls.c:2218:
    Masaru> undefined reference to `gnutls_digest_get_name' gnutls.o: In
    Masaru> function `gnutls_symmetric':
    Masaru> /tmp/tes/emacs/src/gnutls.c:2038: undefined reference to
    Masaru> `gnutls_cipher_get_iv_size'
    Masaru> /tmp/tes/emacs/src/gnutls.c:2055: undefined reference to
    Masaru> `gnutls_cipher_get_tag_size' collect2: error: ld returned 1
    Masaru> exit status Makefile:598: recipe for target 'temacs' failed
    Masaru> make[2]: *** [temacs] Error 1 make[2]: Leaving directory
    Masaru> '/tmp/tes/emacs/src' Makefile:416: recipe for target 'src'
    Masaru> failed make[1]: *** [src] Error 2 make[1]: Leaving directory
    Masaru> '/tmp/tes/emacs' Makefile:1097: recipe for target
    Masaru> 'bootstrap' failed make: *** [bootstrap] Error 2

    Masaru> --- Masaru Nomiya

It may be that you still have effects from 'old files' in you git emacs
distribution. If 'make clean' doesn't help, try a completely new
'git clone git://git.sv.gnu.org/emacs.git'. That's what I did.

Best wishes,

Colin.

-- 
--
Colin Baxter
m43cap@yandex.com
GnuPG fingerprint: 68A8 799C 0230 16E7 BF68  2A27 BBFA 2492 91F5 41C8

Re: gnutls error

[Masaru Nomiya]

Hello,

In the Message; 

  Subject    : Re: gnutls error
  Message-ID : <87o9sles2b.fsf@colin-baxter.com>
  Date & Time: Sat, 15 Jul 2017 14:41:00 +0100

Colin Baxter <colinb@colin-baxter.com> has written:

> Dear Masaru,

> >>>>> "Masaru" == Masaru Nomiya <nomiya@galaxy.dti.ne.jp> writes:

>     Masaru> Hello, In the Message;

>     Masaru>   Subject : Re: gnutls error Message-ID :
>     Masaru> <87shhxex1d.fsf@yandex.com> Date & Time: Sat, 15 Jul 2017
>     Masaru> 12:53:34 +0100

>     Masaru> Colin Baxter <m43cap@yandex.com> has written:

>     >> >>>>> "Eli" == Eli Zaretskii <eliz@gnu.org> writes:

>     Eli> Please try again, I think I fixed this.

>     >> Yes, the error no longer occurs. Emacs-26 now builds
>     >> satisfactorily, for me. Thank you.

>     Masaru> Did you?

>     Masaru> I still got errors with the very latest git head;

>     Masaru> [...]  CCLD temacs gnutls.o: In function `Fgnutls_ciphers':
>     Masaru> /tmp/tes/emacs/src/gnutls.c:1860: undefined reference to
>     Masaru> `gnutls_cipher_get_tag_size'
>     Masaru> /tmp/tes/emacs/src/gnutls.c:1876: undefined reference to
>     Masaru> `gnutls_cipher_get_iv_size' gnutls.o: In function
>     Masaru> `Fgnutls_macs': /tmp/tes/emacs/src/gnutls.c:2197: undefined
>     Masaru> reference to `gnutls_mac_get_nonce_size' gnutls.o: In
>     Masaru> function `Fgnutls_digests':
>     Masaru> /tmp/tes/emacs/src/gnutls.c:2213: undefined reference to
>     Masaru> `gnutls_digest_list' /tmp/tes/emacs/src/gnutls.c:2218:
>     Masaru> undefined reference to `gnutls_digest_get_name' gnutls.o: In
>     Masaru> function `gnutls_symmetric':
>     Masaru> /tmp/tes/emacs/src/gnutls.c:2038: undefined reference to
>     Masaru> `gnutls_cipher_get_iv_size'
>     Masaru> /tmp/tes/emacs/src/gnutls.c:2055: undefined reference to
>     Masaru> `gnutls_cipher_get_tag_size' collect2: error: ld returned 1
>     Masaru> exit status Makefile:598: recipe for target 'temacs' failed
>     Masaru> make[2]: *** [temacs] Error 1 make[2]: Leaving directory
>     Masaru> '/tmp/tes/emacs/src' Makefile:416: recipe for target 'src'
>     Masaru> failed make[1]: *** [src] Error 2 make[1]: Leaving directory
>     Masaru> '/tmp/tes/emacs' Makefile:1097: recipe for target
>     Masaru> 'bootstrap' failed make: *** [bootstrap] Error 2

>     Masaru> --- Masaru Nomiya

> It may be that you still have effects from 'old files' in you git emacs
> distribution. If 'make clean' doesn't help, try a completely new
> 'git clone git://git.sv.gnu.org/emacs.git'. That's what I did.

Yes, I did 'git clone', again, and got the above errors.

Thanks,

---
 Masaru Nomiya

Re: gnutls error

[Eli Zaretskii]

> Date: Sat, 15 Jul 2017 21:42:35 +0900
> From: Masaru Nomiya <nomiya@galaxy.dti.ne.jp>
> Cc: eliz@gnu.org
> 
> > >>>>> "Eli" == Eli Zaretskii <eliz@gnu.org> writes:
> 
> >     Eli> Please try again, I think I fixed this.
> 
> > Yes, the error no longer occurs. Emacs-26 now builds satisfactorily, for
> > me. Thank you.
> 
> Did you?
> 
> I still got errors with the very latest git head;

What version of GnuTLS do you have installed?

Please try the latest master, I've tried to fix this.

Re: gnutls error

[Masaru Nomiya]

Hello,

In the Message; 

  Subject    : Re: gnutls error
  Message-ID : <83wp79erds.fsf@gnu.org>
  Date & Time: Sat, 15 Jul 2017 16:55:43 +0300

Eli Zaretskii <eliz@gnu.org> has written:

> > Date: Sat, 15 Jul 2017 21:42:35 +0900
> > From: Masaru Nomiya <nomiya@galaxy.dti.ne.jp>
> > Cc: eliz@gnu.org
> > 
> > > >>>>> "Eli" == Eli Zaretskii <eliz@gnu.org> writes:
> > 
> > >     Eli> Please try again, I think I fixed this.
> > 
> > > Yes, the error no longer occurs. Emacs-26 now builds satisfactorily, for
> > > me. Thank you.
> > 
> > Did you?
> > 
> > I still got errors with the very latest git head;

> What version of GnuTLS do you have installed?

It's '3.0.28'.

> Please try the latest master, I've tried to fix this.

Sorry, but to no avail.

Thanks,

-----
 Masaru Nomiya

Re: gnutls error

[Eli Zaretskii]

> Date: Sat, 15 Jul 2017 23:07:53 +0900
> From: Masaru Nomiya <nomiya@galaxy.dti.ne.jp>
> CC: emacs-devel@gnu.org
> 
> > What version of GnuTLS do you have installed?
> 
> It's '3.0.28'.
> 
> > Please try the latest master, I've tried to fix this.
> 
> Sorry, but to no avail.

Please show all the error messages you get this time while compiling.
I cannot fix the problem without seeing the errors.

Thanks.

Re: gnutls error

[Dani Moncayo]

On Sat, Jul 15, 2017 at 1:33 PM, Eli Zaretskii <eliz@gnu.org> wrote:
>> From: Dani Moncayo <dmoncayo@gmail.com>
>> Date: Sat, 15 Jul 2017 13:14:35 +0200
>>
>> I've tried to build the current version of the master branch on
>> Ubuntu-GNOME 17.04; it fails this way:
>>
>> [...]
>>   CC       floatfns.o
>>   CC       fns.o
>> In file included from ../../repo/src/fns.c:30:0:
>> ../../repo/src/fns.c:4748:34: error: ‘Fsecure_hash_algorithms’
>> undeclared here (not in a function)
>>  DEFUN ("secure-hash-algorithms", Fsecure_hash_algorithms,
>>                                   ^
>> ../../repo/src/lisp.h:2925:26: note: in definition of macro ‘DEFUN’
>>         { .a ## maxargs = fnname },     \
>>                           ^~~~~~
>> ../../repo/src/fns.c:4748:34: warning: no previous prototype for
>> ‘Fsecure_hash_algorithms’ [-Wmissing-prototypes]
>>  DEFUN ("secure-hash-algorithms", Fsecure_hash_algorithms,
>>                                   ^
>> ../../repo/src/lisp.h:2927:16: note: in definition of macro ‘DEFUN’
>>     Lisp_Object fnname
>>                 ^~~~~~
>> ../../repo/src/fns.c: In function ‘extract_data_from_object’:
>> ../../repo/src/fns.c:4928:24: error: ‘Qiv_auto’ undeclared (first use
>> in this function)
>>    else if (EQ (object, Qiv_auto))
>>                         ^~~~~~~~
>> ../../repo/src/fns.c:4928:24: note: each undeclared identifier is
>
> These errors make no sense, the stuff they complain about is defined
> right there in fns.c.
>
> How did you build Emacs?  With what sequence of commands?

From the root of the source tree:
  $ git pull
  $ ./autogen.sh

From an empty directory:
  $ path-to-source-root/configure
  $ make

(BTW, I've just repeated the above, with the same result)

>  Are you
> sure your Git clone is up to date with upsteam?

I think so. "git log -1" gives here this:
------------------------------------------------
commit b30ee0c9225bad6e3fd0b511a6c5d9a64b8fd66a
Author: Eli Zaretskii <eliz@gnu.org>
Date:   Sat Jul 15 16:54:12 2017 +0300

    Avoid link errors with older versions of GnuTLS

    * src/gnutls.c (Fgnutls_ciphers, gnutls_symmetric_aead)
    (Fgnutls_macs, Fgnutls_digests): Conditionally compile code that
    calls GnuTLS functions which might be unavailable in older
    versions of GnuTLS.
------------------------------------------------

-- 
Dani Moncayo

Re: gnutls error

[Dani Moncayo]

>>  Are you
>> sure your Git clone is up to date with upsteam?
>
> I think so. "git log -1" gives here this:
> ------------------------------------------------
> commit b30ee0c9225bad6e3fd0b511a6c5d9a64b8fd66a
> [...]

Oh, and BTW, "git status" shows that I have no uncommitted local changes:

$ git status
On branch master
Your branch is up-to-date with 'origin/master'.
nothing to commit, working tree clean

-- 
Dani Moncayo

Re: gnutls error

[Masaru Nomiya]

Hello,

In the Message; 

  Subject    : Re: gnutls error
  Message-ID : <83vamtepgh.fsf@gnu.org>
  Date & Time: Sat, 15 Jul 2017 17:37:18 +0300

Eli Zaretskii <eliz@gnu.org> has written:

> > Date: Sat, 15 Jul 2017 23:07:53 +0900
> > From: Masaru Nomiya <nomiya@galaxy.dti.ne.jp>
> > CC: emacs-devel@gnu.org
> > 
> > > What version of GnuTLS do you have installed?
> > 
> > It's '3.0.28'.
> > 
> > > Please try the latest master, I've tried to fix this.
> > 
> > Sorry, but to no avail.

> Please show all the error messages you get this time while compiling.
> I cannot fix the problem without seeing the errors.

According to the Colin's advice, I upgraded gnutls to 3.2.18. This solved
the problem.

Thanks,

---
 Masaru Nomiya
 

Re: gnutls error

[Eli Zaretskii]

> From: Dani Moncayo <dmoncayo@gmail.com>
> Date: Sat, 15 Jul 2017 17:06:52 +0200
> Cc: Emacs development discussions <emacs-devel@gnu.org>
> 
> > These errors make no sense, the stuff they complain about is defined
> > right there in fns.c.
> >
> > How did you build Emacs?  With what sequence of commands?
> 
> >From the root of the source tree:
>   $ git pull
>   $ ./autogen.sh
> 
> >From an empty directory:
>   $ path-to-source-root/configure
>   $ make
> 
> (BTW, I've just repeated the above, with the same result)

What does this produce:

  fgrep Qiv_auto src/globals.h

GnuTLS build-time configuration

[Paul Eggert]

While we're fiddling with GnuTLS configuration, is there any reason we use 
separate configure-time checks for HAVE_GNUTLS3_AEAD, HAVE_GNUTLS3_CIPHER, etc., 
instead of the more-typical approach of using GNUTLS_VERSION_NUMBER? If not, I 
propose simplifying configuration by doing things in the more-typical way. See 
Bug#27708.

https://debbugs.gnu.org/cgi/bugreport.cgi?bug=27708

Re: GnuTLS build-time configuration

[Colin Baxter]

Dear Paul,

>>>>> "Paul" == Paul Eggert <eggert@cs.ucla.edu> writes:

    Paul> While we're fiddling with GnuTLS configuration, is there any
    Paul> reason we use separate configure-time checks for
    Paul> HAVE_GNUTLS3_AEAD, HAVE_GNUTLS3_CIPHER, etc., instead of the
    Paul> more-typical approach of using GNUTLS_VERSION_NUMBER? If not,
    Paul> I propose simplifying configuration by doing things in the
    Paul> more-typical way. See Bug#27708.

    Paul> https://debbugs.gnu.org/cgi/bugreport.cgi?bug=27708

Forgive my amateur (I'm not a cs grad) question, but are you thinking of
using GNUTLS_VERSION_NUMBER to allow GnuTLS versions of say xxx and above?
As in 'if GNUTLS_VERSION_NUMBER > xxx'. I don't know the exact syntax.

Best wishes,

Colin.

-- 
--
Colin Baxter
m43cap@yandex.com
GnuPG fingerprint: 68A8 799C 0230 16E7 BF68  2A27 BBFA 2492 91F5 41C8

Re: GnuTLS build-time configuration

[Ted Zlatanov]

On Sat, 15 Jul 2017 09:17:26 -0700 Paul Eggert <eggert@cs.ucla.edu> wrote: 

PE> While we're fiddling with GnuTLS configuration, is there any reason we use
PE> separate configure-time checks for HAVE_GNUTLS3_AEAD, HAVE_GNUTLS3_CIPHER, etc.,
PE> instead of the more-typical approach of using GNUTLS_VERSION_NUMBER? If not, I
PE> propose simplifying configuration by doing things in the more-typical way. See
PE> Bug#27708.

PE> https://debbugs.gnu.org/cgi/bugreport.cgi?bug=27708

Thank you for the suggestion. I did it that way because those pieces
were not introduced at the same time. Also they may not be available
with the same API in a future version. So I would prefer not to rely
purely on the version.

I won't close the bug report for now. Does anyone else have feedback?
Paul, let me know what you think too.

Thanks
Ted

Re: gnutls error

[Dani Moncayo]

> What does this produce:
>
>   fgrep Qiv_auto src/globals.h

$ fgrep Qiv_auto src/globals.h
#define iQiv_auto 661
DEFINE_LISP_SYMBOL (Qiv_auto)
# define Qiv_auto builtin_lisp_symbol (661)

-- 
Dani Moncayo

Re: GnuTLS build-time configuration

[Paul Eggert]

Colin Baxter wrote:
> are you thinking of
> using GNUTLS_VERSION_NUMBER to allow GnuTLS versions of say xxx and above?

Yes, that's what the patch does, e.g., '0x030400 <= GNUTLS_VERSION_NUMBER'.

Re: GnuTLS build-time configuration

[Paul Eggert]

Ted Zlatanov wrote:
> those pieces
> were not introduced at the same time. Also they may not be available
> with the same API in a future version.

Sure, but I don't see how the 'configure'-time check solves this problem. 
Although some of those pieces were introduced before GnuTLS v3, the 
'configure'-time checks do not look for these pieces in older GnuTLS platforms, 
presumably because their semantics changed or because we're not sure they'll 
really work in these older versions. So although the 'configure'-time check 
might look like it's checking the API, it's really doing a version-number check 
as well, and to some extent it's really the version number that is calling the 
shots.

Normally I'm a fan of the Autoconf Way, where one tests behavior and not version 
numbers. Here, though, we have just one implementation with a well-documented 
relationship between version number and API symbols, along with a 
'configure'-time check that already relies on version numbers, so the Autoconf 
Way is not so useful here.

Re: gnutls error

[Eli Zaretskii]

> From: Dani Moncayo <dmoncayo@gmail.com>
> Date: Sat, 15 Jul 2017 21:58:27 +0200
> Cc: Emacs development discussions <emacs-devel@gnu.org>
> 
> > What does this produce:
> >
> >   fgrep Qiv_auto src/globals.h
> 
> $ fgrep Qiv_auto src/globals.h
> #define iQiv_auto 661
> DEFINE_LISP_SYMBOL (Qiv_auto)
> # define Qiv_auto builtin_lisp_symbol (661)

OK.  Can you show the preprocessed source of fns.c?  (Let me know if
you need help in how to produce it.)

Re: gnutls error

[Dani Moncayo]

[-- Attachment #1: Type: text/plain, Size: 437 bytes --]

> OK.  Can you show the preprocessed source of fns.c?  (Let me know if
> you need help in how to produce it.)

I've tried to produce it, by adding this rule to the file
"build-dir/src/Makefile":

  fns.i: fns.c
   $(AM_V_CC)$(CC) -E $(CPPFLAGS) $(ALL_CFLAGS) $(PROFILING_CFLAGS) $< > $@

and then doing:
  cd build-dir/src
  make fns.i

I'm attaching the preprocessed source.  Let me know if this is not
what you want.

-- 
Dani Moncayo

[-- Attachment #2: fns.i.tar.gz --]
[-- Type: application/x-gzip, Size: 236707 bytes --]

Re: gnutls error

[Dani Moncayo]

> I'm attaching the preprocessed source.  Let me know if this is not
> what you want.

Hi again, Eli.

I've just realized that I had "Makefile"s in the source tree (from an
old in-tree compilation).  So, to clean things up, I've done a "make
distclean" in the top of the source tree, and then I've done a
"configure & make" from an empty build directory.  This time the build
has succeeded.

Sorry for the noise.

-- 
Dani Moncayo

Re: gnutls error

[Eli Zaretskii]

> From: Dani Moncayo <dmoncayo@gmail.com>
> Date: Sun, 16 Jul 2017 18:02:17 +0200
> Cc: Emacs development discussions <emacs-devel@gnu.org>
> 
> > OK.  Can you show the preprocessed source of fns.c?  (Let me know if
> > you need help in how to produce it.)
> 
> I've tried to produce it, by adding this rule to the file
> "build-dir/src/Makefile":
> 
>   fns.i: fns.c
>    $(AM_V_CC)$(CC) -E $(CPPFLAGS) $(ALL_CFLAGS) $(PROFILING_CFLAGS) $< > $@
> 
> and then doing:
>   cd build-dir/src
>   make fns.i
> 
> I'm attaching the preprocessed source.  Let me know if this is not
> what you want.

It is, thanks.

The problem is that you have an outdated globals.h in the source tree,
and the build picks it up instead of the version produced during the
build, which is somewhere in the build tree.  I guess you started by
building in the source tree, and then switched to an out-of-tree
builds, and didn't clean up the source tree.  Or something to that
effect.

Re: GnuTLS build-time configuration

[Ted Zlatanov]

On Sat, 15 Jul 2017 17:16:55 -0700 Paul Eggert <eggert@cs.ucla.edu> wrote: 

PE> Normally I'm a fan of the Autoconf Way, where one tests behavior and not version
PE> numbers. Here, though, we have just one implementation with a well-documented
PE> relationship between version number and API symbols, along with a
PE> 'configure'-time check that already relies on version numbers, so the Autoconf
PE> Way is not so useful here.

I trust your expertise; you have a much better feel for the interaction
of autoconf, the core C code, and external libraries. So please go ahead
and push your change, I'm OK with it.

Thanks for your help
Ted