cannot configure libogg #264
Comments
|
There should be a |
|
Here you go: config.log |
|
This looks like a gcc/clang issue. The failure that makes it abort compiles fine with gcc, but fails with clang. The error is gcc accepts As a workaround, you can modify the configure script to not use |
|
btw, I tried it on mac and while the console output looks the same, the config.log is completely different. I've uploaded it here. |
|
Ah, you're right about the clang issue. Trying directly also shows the error. |
|
Weirdly, using clang for configuring (on unix) works: The config.log is here. The command its using still has O20, but also has O2 (??) which seems to make clang happy: |
|
I've filed a bug with clang here. |
|
@kripken I've been able to work around the -O20 problem by editing the configure file, as you suggested. :) However, on mac its failing for completely different reasons (see config.log). |
|
? I see the same |
|
I think you might be looking at the wrong config.log. Here's a link to it on its [own page](https://raw.github.com/gist/1869963/efbc73cd1c5946df529b5e6e42e054e8845466f8/config.log%20(mac%20os%20x). |
|
Oh ok, sorry. Linker errors in there, it seems? Do you get them on that machine when you run the command directly too? |
|
BTW: On my Ubuntu 11.04 32bit, If I replace all occurrences of |
|
@peterleecomcn GCC technically doesn't support -O4. It instead treats it as -O3. http://stackoverflow.com/questions/3399821/gcc-o4-optimization-flag Clang is less graceful about optimizations above 3, reporting the error mentioned above (error: invalid value '4' in '-O20'). |
|
If the only issue is clang not liking -O4+, i think we can close this. If there are other open things here please comment. |
|
I'm getting the second issue @paleozogt mentioned, on os x. During the configure scripts 'checking size of [data type]...' phase, every test fails with the same error: https://gist.github.com/chrismcr/5490366 |
|
Can you try to configure it with |
|
That gets me an error about /lib/cpp (which doesn't exist) - https://gist.github.com/chrismcr/5490551 |
|
If I force CPP=/usr/bin/cpp, I get the same linker errors as before. I am seeing references to conftest.js in certain places though so it appears EMCONFIGURE_JS is being set. |
|
Ok, can you give me full steps to reproduce and i'll try this locally? |
|
Sure - it's just a straight emconfigure ./configure => "No 16 bit type found on this platform!" Out of interest, I've checked the config.log after setting EMCONFIGURE_JS again, and the "checking size..." tests are definitely still native, when earlier tests are js. |
|
It looks like EMCONFIGURE_JS won't help since the specific test does file io (so we need to use a native executable). I don't know enough about configure scripts to debug this. I can see the failing program in config.log, but not sure why it prints out that error message, looks like the variable for 16-bit is not set. Anyone here an autotools person? ;) |
|
Let me share my experience. config.log showed an invalid value thus, I changed it and... ...it worked. |
GCC and clang accept huge -O* values like -O4 and -O20, silently clamping them to a sensible -O3. Some packages rely on this and specify -O20 as a default "REALLY FAST", such as libogg and libvorbis. For compatibility, accept the large values, clamp them to 3, and print a warning instead of throwing and exception and dying. Fixes emscripten-core#1499 Fixes emscripten-core#264
GCC and clang accept huge -O* values like -O4 and -O20, silently clamping them to a sensible -O3. Some packages rely on this and specify -O20 as a default "REALLY FAST", such as libogg and libvorbis. For compatibility, accept the large values, clamp them to 3, and print a warning instead of throwing an exception and dying. Fixes emscripten-core#1499 Fixes emscripten-core#264
GCC and clang accept huge -O* values like -O4 and -O20, silently clamping them to a sensible -O3. Some packages rely on this and specify -O20 as a default "REALLY FAST", such as libogg and libvorbis. For compatibility, accept the large values, clamp them to 3, and print a warning instead of throwing an exception and dying. Test case added. Fixes emscripten-core#1499 Fixes emscripten-core#264
GCC and clang accept huge -O* values like -O4 and -O20, silently clamping them to a sensible -O3. Some packages rely on this and specify -O20 as a default "REALLY FAST", such as libogg and libvorbis. For compatibility, accept the large values, clamp them to 3, and print a warning instead of throwing an exception and dying. Test case added. Fixes emscripten-core#1499 Fixes emscripten-core#264
I'm trying to build libogg. When I run configure it errors with
No 16 bit type found on this platform!Here's the command and full configure output:
The text was updated successfully, but these errors were encountered: