Patch 7.4.2102

[Bram Moolenaar]

Patch 7.4.2102 (after 7.4.2101)
Problem: Tiny build with GUI fails.
Solution: Revert one FOR_ALL_ change.
Files: src/gui.c


*** ../vim-7.4.2101/src/gui.c 2016-07-24 21:58:39.704057634 +0200
--- src/gui.c 2016-07-24 22:22:54.802631965 +0200
***************
*** 4166,4172 ****
/* avoid that moving components around generates events */
++hold_gui_events;

! FOR_ALL_WINDOWS(wp)
{
if (wp->w_buffer == NULL) /* just in case */
continue;
--- 4166,4172 ----
/* avoid that moving components around generates events */
++hold_gui_events;

! for (wp = firstwin; wp != NULL; wp = W_NEXT(wp))
{
if (wp->w_buffer == NULL) /* just in case */
continue;
*** ../vim-7.4.2101/src/version.c 2016-07-24 21:58:39.720057487 +0200
--- src/version.c 2016-07-24 22:24:22.025828318 +0200
***************
*** 760,761 ****
--- 760,763 ----
{ /* Add new patch number below this line */
+ /**/
+ 2102,
/**/

--
If cars evolved at the same rate as computers have, they'd cost five euro,
run for a year on a couple of liters of petrol, and explode once a day.

/// Bram Moolenaar -- Br...@Moolenaar.net -- http://www.Moolenaar.net \\\
/// sponsor Vim, vote for features -- http://www.Vim.org/sponsor/ \\\
\\\ an exciting new programming language -- http://www.Zimbu.org ///
\\\ help me help AIDS victims -- http://ICCF-Holland.org ///

[Hisashi T Fujinaka]

Well, --huge is breaking on OSX for me again. I'm seeing if I can figure
out where it broke. I wish I knew how to use bisect...

--
Hisashi T Fujinaka - ht...@twofifty.com
BSEE + BSChem + BAEnglish + MSCS + $2.50 = coffee

[Tony Mechelynck]

On Mon, Jul 25, 2016 at 7:41 PM, Hisashi T Fujinaka <ht...@twofifty.com> wrote:
> On Sun, 24 Jul 2016, Bram Moolenaar wrote:
>
>>
>> Patch 7.4.2102 (after 7.4.2101)
>> Problem: Tiny build with GUI fails.
>> Solution: Revert one FOR_ALL_ change.
>> Files: src/gui.c
>

> Well, --huge is breaking on OSX for me again. I'm seeing if I can figure
> out where it broke. I wish I knew how to use bisect...
>

Well, FWIW, it is not breaking for me on Linux64 with GTK2/Gnome GUI
so the problem must be at some point where OSX and Linux are handled
differently.

The principle of bisection is to mark a "last known good" and a
"first known bad" version; then you compile some version approximately
halfway between them and test it. If it is "good" you have a new,
later, "last known good". If it is bad, you have a new, earlier,
"first known bad". Then you repeat with successively narrower
intervals until you have a "known good" immediately followed by a
"known bad" -- which is the actual "first bad ever" where the error
was made.


Best regards,
Tony.

[Manuel Ortega]

On Mon, Jul 25, 2016 at 1:41 PM, Hisashi T Fujinaka <ht...@twofifty.com> wrote:

Well, --huge is breaking on OSX for me again. I'm seeing if I can figure
out where it broke. I wish I knew how to use bisect...

What are your ./configure flags?  I am building with huge features on OSX with no problems.

-Manny

[Manuel Ortega]

By the way, --huge is the default now, so you shouldn't need to pass it at the command line.

Bram: "./configure --help" still claims that the default level is "normal".

-Manny 

[Hisashi T Fujinaka]

Um, no. I meant I don't know how to get it to work automatically.

[Hisashi T Fujinaka]

I'm doing:

make clean && ./configure --with-features=huge && make test

[Hisashi T Fujinaka]

On Mon, 25 Jul 2016, Hisashi T Fujinaka wrote:

> On Mon, 25 Jul 2016, Manuel Ortega wrote:
>
>> On Mon, Jul 25, 2016 at 2:00 PM, Manuel Ortega <manny...@gmail.com>
>> wrote:
>>
>>
>> On Mon, Jul 25, 2016 at 1:41 PM, Hisashi T Fujinaka
>> <ht...@twofifty.com> wrote:
>>
>> Well, --huge is breaking on OSX for me again. I'm seeing if I
>> can figure
>> out where it broke. I wish I knew how to use bisect...
>>
>>
>> What are your ./configure flags? I am building with huge features on OSX
>> with no problems.
>>
>>
>> By the way, --huge is the default now, so you shouldn't need to pass it at
>> the command line.
>>
>> Bram: "./configure --help" still claims that the default level is "normal".
>
> I'm doing:
>
> make clean && ./configure --with-features=huge && make test

I must have something else broken because I'm all the way back to
v7.4.2054 and it's still not building.

--

[Tony Mechelynck]

On Mon, Jul 25, 2016 at 9:27 PM, Hisashi T Fujinaka <ht...@twofifty.com> wrote:
> On Mon, 25 Jul 2016, Manuel Ortega wrote:
>
>> On Mon, Jul 25, 2016 at 2:00 PM, Manuel Ortega <manny...@gmail.com>
>> wrote:
>>
>>
>> On Mon, Jul 25, 2016 at 1:41 PM, Hisashi T Fujinaka
>> <ht...@twofifty.com> wrote:
>>
>> Well, --huge is breaking on OSX for me again. I'm seeing if I
>> can figure
>> out where it broke. I wish I knew how to use bisect...
>>
>>
>> What are your ./configure flags? I am building with huge features on OSX
>> with no problems.
>>
>>
>> By the way, --huge is the default now, so you shouldn't need to pass it at
>> the command line.
>>
>> Bram: "./configure --help" still claims that the default level is
>> "normal".
>
>
> I'm doing:
>
> make clean && ./configure --with-features=huge && make test

I'm not using "make test". I set my configure setting by means of
environment variables (so they will not be forgotten if make decides
to run configure) then I run just "make" (or "make reconfig" if I have
changed the settings or if some important libray package has been
upgraded) followed by "make install" for my Huge build or by "make
installruntime" for my Tiny build, which has --with-vim-name=vi to set
it apart from the Huge build.

Then if completion and link are OK I use mostly the Huge (gui-enabled)
version as either gvim or vim. I haven't had any problems so far with
7.4.2012 since it was published.

Is your problem with "compiling" Vim (then which messages do you
get?), with "using" it yourself (and what are the symptoms?) or with
running "make test"? If the latter, there is the possibility that the
failing test (which one and how?) is the problem.


Best regards,
Tony.

[Tony Mechelynck]

Please post the error messages you get.

Does it still not build if instead of "make clean" you use "make distclean"?


Best regards,
Tony.

[Hisashi T Fujinaka]

Sorry, but lately most of my comments have been about bad tests. I
assumed most people read all the email here but that was a bad
assumption.

Bram broke my builds a while ago (like in the 203x range) in netbseans
and the tests yield:

From test_netbeans.vim:
Found errors in Test_nb_basic():
function RunTheTest[9]..Test_nb_basic[2]..<SNR>4_run_server[1]..RunServer[21]..Nb_basic line 11: Expected 2 but got 130
function RunTheTest[9]..Test_nb_basic[2]..<SNR>4_run_server[1]..RunServer[21]..Nb_basic line 12: Expected 20 but got 2
function RunTheTest[9]..Test_nb_basic[2]..<SNR>4_run_server[1]..RunServer[21]..Nb_basic line 23: Expected '0:disconnect=1' but got '0:disconnect=0'

Test results:


From test_netbeans.vim:
Found errors in Test_nb_basic():
function RunTheTest[9]..Test_nb_basic[2]..<SNR>4_run_server[1]..RunServer[21]..Nb_basic line 11: Expected 2 but got 130
function RunTheTest[9]..Test_nb_basic[2]..<SNR>4_run_server[1]..RunServer[21]..Nb_basic line 12: Expected 20 but got 2
function RunTheTest[9]..Test_nb_basic[2]..<SNR>4_run_server[1]..RunServer[21]..Nb_basic line 23: Expected '0:disconnect=1' but got '0:disconnect=0'
TEST FAILURE

I asked Bram about this (and about whether netbeans was even supposed to
be used in OSX) and all he said was "works for me". Does anyone else
even have Java installed?

--

[Bram Moolenaar]

Dominique discovered that it also breaks when running under Valgrind.
I suspect a race condition.

--
A hamburger walks into a bar, and the bartender says: "I'm sorry,
but we don't serve food here."

[Hisashi T Fujinaka]

That would explain why it works for me sometimes.