Android builds broken by datafiles.c

[bob-the-hamster]

Android builds have been broken by the addition of datafiles.c

jni/../jni/application/ohrrpgce/tmp//filelayer.cpp:702: error: undefined reference to 'embedded_files_table'
collect2: error: ld returned 1 exit status
/home/james/misc/android-ndk-r12b/build/core/build-binary.mk:677: recipe for target 'obj/local/armeabi/libapplication.so' failed
make: *** [obj/local/armeabi/libapplication.so] Error 1

tmc correctly points out:

Oops. Probably because datafiles.c is generate under build/ so doesn't
get symlinked into android/tmp.

A quick hack to fix that does avoid the first error message:

Index: ohrbuild.py
===================================================================
--- ohrbuild.py	(revision 12602)
+++ ohrbuild.py	(working copy)
@@ -495,7 +495,12 @@
             source_nodes += [node]
         else:
             # node.sources[0] itself is a path in build/ (to a nonexistent file)
-            source_files.append (node.sources[0].srcnode().abspath)
+            if node.sources[0].name.endswith('datafiles.c'):
+                # special case for datafiles.c which is created in builds dir
+                p = node.sources[0].srcnode().abspath
+                source_files.append (os.path.join(os.path.dirname(p), "build", os.path.basename(p)))
+            else:
+                source_files.append (node.sources[0].srcnode().abspath)
             source_nodes += node.sources
 
     # hacky. Copy the right source files to a temp directory because the Android.mk used

However, then compiling for android results in a colossal cascade of errors like this:

jni/../jni/application/ohrrpgce/tmp//datafiles.c:4:3: error: 'R' undeclared here (not in a function)
   R,E,L,D,^A,^M, , , ,<A6>,^G, , ,<95>,^G, , , , ,^F,^E, , , ,^A,^B,@,^A, ,^E, , , ,^B,^B,<C8>, , ,^D, ,
   ^
jni/../jni/application/ohrrpgce/tmp//datafiles.c:4:5: error: 'E' undeclared here (not in a function)
   R,E,L,D,^A,^M, , , ,<A6>,^G, , ,<95>,^G, , , , ,^F,^E, , , ,^A,^B,@,^A, ,^E, , , ,^B,^B,<C8>, , ,^D, ,
     ^
jni/../jni/application/ohrrpgce/tmp//datafiles.c:4:7: error: 'L' undeclared here (not in a function)
   R,E,L,D,^A,^M, , , ,<A6>,^G, , ,<95>,^G, , , , ,^F,^E, , , ,^A,^B,@,^A, ,^E, , , ,^B,^B,<C8>, , ,^D, ,
       ^
jni/../jni/application/ohrrpgce/tmp//datafiles.c:4:9: error: 'D' undeclared here (not in a function)
   R,E,L,D,^A,^M, , , ,<A6>,^G, , ,<95>,^G, , , , ,^F,^E, , , ,^A,^B,@,^A, ,^E, , , ,^B,^B,<C8>, , ,^D, ,
         ^
jni/../jni/application/ohrrpgce/tmp//datafiles.c:4:3: error: stray '\1' in program
   R,E,L,D,^A,^M, , , ,<A6>,^G, , ,<95>,^G, , , , ,^F,^E, , , ,^A,^B,@,^A, ,^E, , , ,^B,^B,<C8>, , ,^D, ,
   ^
jni/../jni/application/ohrrpgce/tmp//datafiles.c:4:12: error: expected expression before ',' token
   R,E,L,D,^A,^M, , , ,<A6>,^G, , ,<95>,^G, , , , ,^F,^E, , , ,^A,^B,@,^A, ,^E, , , ,^B,^B,<C8>, , ,^D, ,
            ^
jni/../jni/application/ohrrpgce/tmp//datafiles.c:5:2: warning: null character(s) ignored
    , ,^C,^A,<FF>, ,^D, , , ,^D,^A,<FF>, ,^M, , , ,^E,^F, ,C,o,n,t,a,i,n,e,r, ,[,^G, , ,^F, ,^B,*,^F,

[bob-the-hamster]

Re-opening. The last commit fixed the problem where the Android build had data files.c symlinked wrong, but the "undeclared here (not in a function)" errors are still happening

[rversteegen]

It compiles fine for me. I think I misinterpreted the error you posted as being that you had managed to make datafiles.c a symlink to one of the sourceslice files, but now it looks like actually generate_datafiles_c is broken, not converting the bytes of the input to integers. But why does it only break when you attempt to build for android? Probably it's the python version or $LANG you are using when doing that.

        with open(path, 'rb') as datafile:
            data = datafile.read()
            ...
            for offset in range(0, len(data), 40):
                row = data[offset : offset + 40]
                ret += '  ' + ','.join(str(byte) for byte in row) + ',\n'

The problem seems to be str

[rversteegen]

Should be fixed now although I haven't gotten the commit email (I may have overloaded the mail server!); I tested with Python 2.