2010-07-13 06:23:52     mpd compile problem

Document created by Aaronwu Employee on Aug 22, 2013
Version 1Show Document
  • View in full screen mode

2010-07-13 06:23:52     mpd compile problem

Rob Maris (GERMANY)

Message: 91194   

 

Referring to thread 82339, I'm obviously encountering assembler problems, because a false include path is specified with the make process. However, I'm working with SVN trunk mpd-0.13.2, where autoconfiguration should do well (I expect).

 

Here relevant portion of make output:

 

bfin-linux-uclibc-gcc -DHAVE_CONFIG_H -I. -I../../mpd-0.13.2/src -I.. -DEMBED -D__uClinux__ \

    -I/home/rob/uClinux_2.6.4/trunk -mcpu=bf537-0.2 -Wall -Wmissing-prototypes -I/usr/include \

    -pipe -Wall -g -O2 -mcpu=bf537-0.2 -c -o mpd-inputStream_http.o `test -f 'inputStream_http.c' \

    || echo '../../mpd-0.13.2/src/'`inputStream_http.c

{standard input}: Assembler messages:

{standard input}:1412: Error: syntax error. Input text was cld.

{standard input}:1412: Error:

{standard input}:1412: Error: syntax error. Input text was rep.

 

Apparently, the -I/usr/include represents the problem. On the other hand: hard to imagine that an error exists in the existent autoconfigure process, since mpd is part of a nominally correctly operating netaudio app - BTW: I'm not using a stamp board, but instead a bluetechnix tcm-bf537.

 

Current problems arose after continuing my mpd testing effort (refer to my posting 89216 - where audio stream is reported not to operate): I found that libao was not included automatically by activating NETAUDIO in menuconfig (according to the netaudio docs page). After activating this, the assembler errors occur. Prior to activating this, I got the streaming problems. Since no compile errors are generated when libao is not found, this is quite causing a bit confusion regarding the probable cause of that problem. If anybody can give a hint regarding such error - welcome. I'd wish to learn about how to get the correct conclusions when an app does not work while apparently compiling without errors (in this case: when libao not included).

 

-Rob

 

P.S.: I'd suggest to add appended file to mpd's directory

 

README_UCLINUX

QuoteReplyEditDelete

 

 

2010-07-13 14:24:49     Re: mpd compile problem

Mike Frysinger (UNITED STATES)

Message: 91208   

 

look at the config.log and Makefile to figure out where exactly the -I/usr/include is coming from

QuoteReplyEditDelete

 

 

2010-07-14 05:14:11     Re: mpd compile problem

Rob Maris (GERMANY)

Message: 91236   

 

I'm sorry, but I don' t see the route to solve the problem. When I temporarily modify AO_CFLAGS and MPD_CFLAGS into /opt/uClinux/bfin-uclinux/bfin-uclinux/runtime/usr/include, nothing changes (compiler output still with -I/usr/include).

 

When libao is not active, both variables are empty. I have created a diff of log files with and without libao - attached.

 

1. -I/usr/include only present at many places in the libao version.

 

2. "skipping incompatible ...." are a logical consequence.

 

3. in both files, some errors "no such file..." occur, apparantly not a problem.

 

4. related to libmad, a fork error is reported (simply remarke here).

 

Please provide further suggestions on how to proceed.

 

-Rob

 

 

 

 

 

 

 

 

diff_libao

QuoteReplyEditDelete

 

 

2010-07-14 15:39:11     Re: mpd compile problem

Mike Frysinger (UNITED STATES)

Message: 91251   

 

look at the full config.log.  -I/usr/include shows up when looking for id3tag.  so then look at the configure script and the answer should be clear:

if test x$enable_id3 = xyes; then

    if test "x$id3tag_libraries" != "x" ; then

        ID3TAG_LIBS="-L$id3tag_libraries"

    elif test "x$id3tag_prefix" != "x" ; then

        ID3TAG_LIBS="-L$id3tag_prefix/lib"

    fi

 

that looks broken

QuoteReplyEditDelete

 

 

2010-07-14 17:13:29     Re: mpd compile problem

Rob Maris (GERMANY)

Message: 91255   

 

I'm sorry. I've tried to investigate this. Apparently configure scripts are man-made scripts (hm, that huge size of file? inside it is noted that configure is "Generated by GNU Autoconf 2.61").

 

OK, let's check it.

 

Should $id3tag_prefix not be "/usr", but instead /opt/uClinux/......

         or ....... should $id3tag_libraries be an empty string?

 

$id3tag_libraries is assigned with $withval (or empty), which is defined by $with_id3tag, which is apparently defined outside of the configure script.

 

The same applies to $id3tag_prefix.

 

I'm absolutely overstrained with this sort of detail problems. Therefore my questions:

- Do I really need libao in order to have mpd playing audio correctly?

- May any bug have been introduced after demo introduction of netaudio for blackfin by 2005?

- And, following this: is it known by experience, that mpd is operating correclty in recent trunk history, too?

 

Thanks further for any help/assistance.

QuoteReplyEditDelete

 

 

2010-07-14 17:31:07     Re: mpd compile problem

Mike Frysinger (UNITED STATES)

Message: 91256   

 

prefix should be /usr.  the code that assumes $prefix/ is a usable path is broken.

 

i dont know how mpd uses libao.  i imagine the mpd website documents it.

QuoteReplyEditDelete

 

 

2010-07-14 17:57:32     Re: mpd compile problem

Rob Maris (GERMANY)

Message: 91257   

 

Understood. Apparently the code itself does not take cross-compiled target into account. At this time I'd suggest that libao must NOT be checked as is stated here:

 

https://docs.blackfin.uclinux.org/doku.php?id=network_audio_player&s[]=netaudio

(note: libao is not automatically forced like the others, but I simply tested what happens when selected)

 

 

 

I verified the mpd website. It seems that libao is needed only when I specify for audio output "type" the value "ao". Well, my value is "alsa", as outlined here (excerpt from a modified mpd.conf) :

 

audio_output {

type            "alsa"

name            "My ALSA Device"

device          "hw:0,0"        # optional

format          "16000:16:1"    # optional

mixer_device    "default"       # optional

mixer_control   "PCM"           # optional

mixer_index     "0"             # optional

 

(further contributions to be expected in a separete thread without libao assumed)

QuoteReplyEditDelete

 

 

2010-07-31 02:25:12     Re: mpd compile problem

Mike Frysinger (UNITED STATES)

Message: 91945   

 

that wasnt the broken logic in this case.  once again, the XIPH m4 code is terrible and the XIPH_PATH_AO was adding tis ****.  ive fixed it in trunk rev 9769.

QuoteReplyEditDelete

 

 

2010-08-02 06:37:14     Re: mpd compile problem

Rob Maris (GERMANY)

Message: 92045   

 

build with libao now  OK. At this time I still cannot confirm if libao is needed, since my audio with trunk head still not working. I'll return here when confirmation about libao can be made.

Attachments

Outcomes