3.5-nightly-compile/link error. (Linux)

[Spacebux]

Anyone able to point me and my compiler in the correct direction?

[mark@mb1 angband-b1011f7cb2d409b7581ab6ad7fede9bfeb70ff77]$ make
Entering directory src.
cat: ../version: No such file or directory
cat: ../version: No such file or directory
Entering directory monster.
Leaving directory monster.
Entering directory object.
Leaving directory object.
Entering directory player.
Leaving directory player.
cat: ../version: No such file or directory
cat: ../version: No such file or directory
main.c:43: warning: ISO C forbids empty initializer braces
main.c:42: error: zero or negative size array ‘modules’
Failed to compile main.c!
make[3]: *** [main.o] Error 1
make[2]: *** [all] Error 2
make[1]: *** [subdirs] Error 2
make: *** [all] Error 2

I'm not an expert at linker stuff. Where / how do I go about diagnosing this, please?

Linux mb1 2.6.32-358.el6.i686 #1 SMP Thu Feb 21 21:50:49 UTC 2013 i686 i686 i386 GNU/Linux

Thanks.

[Magnate]

Sounds like someone has broken the automated version-generation script. This is pretty arcane, and well beyond me. I will see if noz is around to take a look. If I'm right then master should be broken for everyone on Linux? (I'm not sure if the script is used in the Windows builds.)

But in fact b1101f7 built correctly under the autobuilders, so it looks like it's just you. Do you have git installed? If you just d/l'd the source that might be the problem. I think ./version is generated by a call to git. Odd that you don't get an error message about that though.

[Derakon]

My best guess is that you used the wrong makefile. The file the error occurred in defines a "modules" list that depends on which of certain constants has been defined; the compiler is complaining that the list is empty (i.e. none of the constants exists). Try doing e.g. "make -f Makefile.std".

[Spacebux]

makefile

Thank you Magnate, Derekon for responding so quickly.

I have no Makefile.std file, however.

Yes, Magnate, I just d/l'd the source. I'll re-try with git this time.

Are there any options I should know about when using git?

[Spacebux]

ok - I don't even know how to use git.


git clone git://github.com/angband/angband.git /tmp/angband/angband-dev

that got me something, not sure which version it is, but the 3.4 code I had earlier compiled without issue... so, I think this is the latest 3.5-dev branch.

[Spacebux]

ok, I got something.. not sure which version it is yet..


...

and, it failed with nearly the same error.

[root@mb1 angband-dev]# make
Entering directory src.
Entering directory monster.
Leaving directory monster.
Entering directory object.
Leaving directory object.
Entering directory player.
Leaving directory player.
main.c:43: warning: ISO C forbids empty initializer braces
main.c:42: error: zero or negative size array ‘modules’
Failed to compile main.c!
make[3]: *** [main.o] Error 1
make[2]: *** [all] Error 2
make[1]: *** [subdirs] Error 2
make: *** [all] Error 2
[root@mb1 angband-dev]#

the error seems to have been in the LINK angband.o statement.

[Derakon]

Where did you download the source from? This page is the official source; I wouldn't be surprised if there are other, less-official places you could get older versions from. Install Git, then do "git clone https://github.com/angband/angband.git", and you should get a proper copy.

[Spacebux]

Where did you download the source from? This page is the official source; I wouldn't be surprised if there are other, less-official places you could get older versions from. Install Git, then do "git clone https://github.com/angband/angband.git", and you should get a proper copy.

Yeah, sorry, I edited my earlier post - that's the command I used, except I specified a target location.

[Spacebux]

[root@mb1 angband-dev]# gcc --version
gcc (GCC) 4.4.7 20120313 (Red Hat 4.4.7-3)

[root@mb1 angband-dev]# make --version
GNU Make 3.81

This program built for i386-redhat-linux-gnu

[root@mb1 angband-dev]# ln --version
ln (GNU coreutils) 8.4

[Derakon]

Yeah, sorry, I edited my earlier post - that's the command I used, except I specified a target location.

Right, that should get you the latest 3.5 development version, as you say.

The compiler output you pasted earlier indicates a problem with compiling main.c, not a problem with the linker. Hence why I suggested making certain that you were using the correct Makefile; it seemed like whatever you were using was failing to define at least one important constant. I don't know if Makefile.std is necessarily the one to use, but the comments at the top indicate it's the right one for Linux. You'd use Makefile.osx for Macs, of course, and Makefile.win for Windows (dunno if that's cygwin/mingw or Visual Studio or what).

[Spacebux]

I suggested making certain that you were using the correct Makefile; it seemed like whatever you were using was failing to define at least one important constant. I don't know if Makefile.std is necessarily the one to use, but the comments at the top indicate it's the right one for Linux. You'd use Makefile.osx for Macs, of course, and Makefile.win for Windows (dunno if that's cygwin/mingw or Visual Studio or what).

I gave Makefile.std a whirl - crash & burn... worse than the first one.


Guess I have to keep playing on Windoze for the meantime...

[Spacebux]

Success.

ok, to resolve, I simply added in a number to the empty array in main.c.


Old main.c:42: static const struct module modules[] =


New main.c:42: static const struct module modules[8] =


I still have the warning about line 43, ISO C forbids empty initializer braces
but, compiled all the way through. Yay.

[Spacebux]

... almost.

... and, then, make install tripped up in the documentation phase. (I guess I'm the only one still grappling with Linux thesedays.)

make install ...
.
.
.
Entering directory doc.
/bin/sh: rst2html: command not found
Failed to compile ../lib/help/index.txt!

--- Installed python-docutils package ---

Much better. Might want to have "Need python-docutils package installed" listed somewhere in the compile.txt file.

[Spacebux]

Woohoo!

Seg fault.

Maybe the ISO C warning is more than just a warning....

[Derakon]

Yeah, I think that patching that modules list by just inserting a dummy variable wasn't the right idea. Sure it makes the compile error go away, but it doesn't fix the underlying problem that that error was telling you about. You really need to have at least one of USE_X11, USE_SDL, or USE_GCU defined or else the game won't know how to run.