Compilation with clang for 32bit fails check-defs.sh and check-symbols.sh tests

[moxian]

Steps to reproduce:

$ ./configure --build=i686-pc-linux-gnu --host=i686-pc-linux-gnu  CC='clang -m32' CXX='clang++ -m32' && make && make check
checking for a BSD-compatible install... /usr/bin/install -c
checking whether build environment is sane... yes

<..snip...>

make  check-TESTS
make[4]: Entering directory '/home/moxian/work/harfbuzz/build/src'
make[5]: Entering directory '/home/moxian/work/harfbuzz/build/src'
PASS: check-c-linkage-decls.sh
FAIL: check-defs.sh
PASS: check-header-guards.sh
PASS: check-includes.sh
PASS: check-libstdc++.sh
PASS: check-static-inits.sh
FAIL: check-symbols.sh
PASS: test-ot-tag
============================================================================
Testsuite summary for HarfBuzz 1.4.6
============================================================================
# TOTAL: 8
# PASS:  6
# SKIP:  0
# XFAIL: 0
# FAIL:  2
# XPASS: 0
# ERROR: 0
============================================================================
See src/test-suite.log
Please report to https://github.com/behdad/harfbuzz/issues/new
============================================================================

The contents of src/check-symbols.sh.log (src/check-defs.sh.log is very similar):

Checking that we are not exposing internal symbols
Processing .libs/libharfbuzz.so

<...~35800 blank lines ommited ...>

Ouch, internal symbols exposed
FAIL check-symbols.sh (exit status: 1)

The reason for that is that nm src/.libs/libharfbuzz.so outputs the following:

00000000 N 
0000002d N 
0000004b N 
00000070 N 
00019e85 N 
0000007d N 
<.. ~38000 of similar lines skipped...>
0006139a N 
0000dc71 N 
         U __assert_fail@@GLIBC_2.0
00052730 t atexit
0008663e r bracket_pairs
00094310 B __bss_start
         U calloc@@GLIBC_2.0
<...the rest of "normal" output...>
00074030 r _ZZL14thai_pua_shapej13thai_action_tP9hb_font_tE11SD_mappings
000740e4 r _ZZL14thai_pua_shapej13thai_action_tP9hb_font_tE11SL_mappings
0007409c r _ZZL14thai_pua_shapej13thai_action_tP9hb_font_tE12SDL_mappings

I think empty lines should not be considered to be "exposing internal symbols", and should be filtered for the purpose of the test (e.g. by adding | grep -v '^$' to the EXPORTED_SYMBOLS definition).

[behdad]

I think empty lines should not be considered to be "exposing internal symbols",

I don't understand how that is possible. Both checks already filter grep ' [BCDGINRSTVW] '. How is that matching empty lines? Please debug.

[moxian]

The initial output of nm src/.libs/libharfbuzz.so does contain, say, the following line:

0000002d N 

This is a nonempty line, and it does contain ' N ', which is why this line is not dropped by grep ' [BCDGINRSTVW] '. However this line does not contain any meaningful symbols. When symbol names are being "extracted" from it (by chosing the third column), the symbol name turns out to be empty string.

[behdad]

Got it. Thanks.

[moxian]

Can confirm it's fixed now.