freetype_versioncheck failed to compile
kelly.ohair at oracle.com
Thu Jan 5 20:46:23 UTC 2012
I got a little carried away... but here is what I came up with:
Just got rid of CC_PROGRAM_OUTPUT_FLAG and also fixed a few indentation annoyances.
The CC_OBJECT_OUTPUT_FLAG is used quite a bit in other places, but
the CC_PROGRAM_OUTPUT_FLAG variable was not used consistently.
Windows has multiple ways to specify the names. So I just got rid of it, everyone is explicit now.
On Jan 4, 2012, at 6:50 PM, Jonathan Lu wrote:
> Hi Kelly,
> Thanks for reviewing,
> On 01/05/2012 06:35 AM, Kelly O'Hair wrote:
>> The change sounds reasonable, but it's a change to something I have always hated, so it's somewhat distasteful to me
>> because of that. Having the makefiles build and run an application as part of a sanity check just seems so...
>> what is the word.... silly? :^(
>> I had hoped that we could just have the sanity check inspect the freetype headers and libraries to
>> insure the right version, not have to build an application just so we could run it to get the version number.
>> On the other hand, building this little app is a way to verify that the freetype library links ok
> Agree, have you got any good ideas about inspecting the headers and libraries? especially for the integrity of a binary library, 'nm libaaa' ?
>> So to the question of whether this change is ok, basically yes, but why was this line added:
>> 53 CC_PROGRAM_OUTPUT_FLAG= -o
> This line is added because there may not be a definition of CC_PROGRAM_OUTPUT_FLAG in jdk/make/common/Defs-<platform>.gmk for all Unix's, so this line will make the little application pass the compilation even without a Defs-<platform>.gmk.
> And if this change is OK, do you plan to push it?
More information about the build-dev