This is the mail archive of the
gcc-help@gcc.gnu.org
mailing list for the GCC project.
Re: Android Native GCC 4.9.2 Build Fails at Dynamic libgcc
- From: Cyd Haselton <chaselton at gmail dot com>
- To: Andrew Haley <aph at redhat dot com>
- Cc: "gcc-help at gcc dot gnu dot org" <gcc-help at gcc dot gnu dot org>
- Date: Fri, 2 Jan 2015 11:38:46 -0600
- Subject: Re: Android Native GCC 4.9.2 Build Fails at Dynamic libgcc
- Authentication-results: sourceware.org; auth=none
- References: <CAHu5PrboyVq7orESm4N3m+2aQZaQNPAkcG4AJ7rFnFu+Q=tKWQ at mail dot gmail dot com> <549A821F dot 1080208 at redhat dot com> <CAHu5PrYKd+-KkODuh0_VMuV03b4H+Lg1rMPUNJPX7CcecywxSA at mail dot gmail dot com> <CAHu5PrYBVudZnMnEfLS3T7e9tRreXFBn6f_gEvgAXmpDYSJCiw at mail dot gmail dot com> <549BCA5C dot 6080708 at redhat dot com> <CAHu5PrY0MUt--KbNCCbQ+4z8jwVHTVaRrE9d+TBwubs3dTzP5Q at mail dot gmail dot com> <CAHu5PrZ1QVLAPyexbqpSafcRuepVfxZdHv3avgFVqXKubC0jug at mail dot gmail dot com> <CAHu5PrYWjwmC7VtqsNSwezvYVvFcuj6EB9UUP_=RKxau5t+kNQ at mail dot gmail dot com> <549DDADB dot 4020707 at redhat dot com> <CAHu5PraJcUxrKcs36uc=njbbs7HLRr4+ozL04OzMEPVyZCDBUw at mail dot gmail dot com> <549EEEF6 dot 9030506 at redhat dot com> <CAHu5PrYG50rU-o7MrFkehdNpQ-KvXYmh6Ymr=OzmYe++V_Wppg at mail dot gmail dot com> <CAHu5PrZ2-Ts7SOqtj=ZdyFbAWpYhf5uEqMAieuJx+1SpmvV1rQ at mail dot gmail dot com> <54A660C4 dot 3030108 at redhat dot com> <CAHu5PrbjFyy83xYUdx+vSVQuVJhD-_HzEAb735o_K+dVbXDLag at mail dot gmail dot com> <54A67520 dot 6030801 at redhat dot com> <CAHu5PrZpsNYawNMacyuZMb9owEsRFf8NN6KqsZZ-zbxYf8_JrA at mail dot gmail dot com> <54A68729 dot 6080004 at redhat dot com>
On Fri, Jan 2, 2015 at 5:55 AM, Andrew Haley <aph@redhat.com> wrote:
> On 01/02/2015 11:48 AM, Cyd Haselton wrote:
>> On Fri, Jan 2, 2015 at 4:38 AM, Andrew Haley <aph@redhat.com> wrote:
>>> On 01/02/2015 10:32 AM, Cyd Haselton wrote:
>>>
>>> Okay, so if the commands are identical you need to run them by hand in
>>> the correct directory to see what is different. if the command which
>>> failed in 4.9 is exactly the same as the command which succeeded
>>> in 4.8 its input files must be different, and in particular dlopen()
>>> must be defined in one of the input files or the user of dlopen()
>>> must not be there in 4.8.
>>>
>> Thanks to the inability to write to extednal storage there is not much
>> space to work with on my android device. Keeping both the 4.8.4 and
>> the 4.9.2 builds plus the build GCC on device eats up nearly all of
>> it.
>
> Sure, but it never even occurred to me that having gone this far you'd
> delete the build tree! That's the problem with trying to debug
> problems like this by mail. If you can login to the device then you
> can copy a tree from it, I would have thought.
I deleted it because, in the previous email I'd asked if I should
build both and compare both directories...you said the following:
>> No. Keep a log of the entire build:
>>
>> make | tee make.log
>>
>> and look at how the dynamic libgcc is linked. Look at the makefile
>> to see where that lib comes from. Look at the 4.9 makefile to see
>> where it is different.
>From that information I determined that the log file was the important
part of the troubleshooting process and...having little space on
device...deleted the build directory.
>
>> That being said...
>>
>> I still have the build logs and therefore the commands for the libgcc
>> link, but the list of input files is huge. I can review them, but if I
>> need to post them here should I use a link to a file or just copy and
>> paste everything?
>
> Neither. Sorry, I think you need to be able to run the commands by
> hand. I can't think of any other way: we really need to be able to
> compare the working and non-working trees.
>
> Andrew.
Okay. I'll re-download 4.8.4 and set up to build...again.
For the record, I copied and pasted the libgcc specific commands and
the giant list of objects from the 4.8 and 4.9 build logs into four
separate files to do a side by side compare of the commands AND the
object files used in said commands. The commands are exactly the same.
The list of object files look exactly the same but, again, the list is
huge so I may be overlooking something.