This is the mail archive of the cygwin-developers@cygwin.com mailing list for the Cygwin project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]

Re: cygwin build SEGV


Christopher Faylor wrote:

> On Wed, Sep 05, 2001 at 12:29:58AM -0400, Charles S. Wilson wrote:
> 
>>
>>Christopher Faylor wrote:
>>
>>
>>>On Tue, Sep 04, 2001 at 11:40:03PM -0400, Christopher Faylor wrote:
>>>
>>>
>>>>[Note that I've renamed the subject]
>>>>
>>>>
>>>Chuck, could you rebuild newlib with -g?  It looks like it was compiled
>>>without this, AFAICT.  This is part of what is screwing up the bt.
>>>
>>
>>Sure.  Coming up.
>>
> 
> Another thing to check:
> 
> frame 2 (func_shell)
> p maxlen
> p i
> p cc
> p buf


I can't.  I had to shut down gdb in order to rebuild cygwin/newlib with 
-g.  I fully expected that at some point during that process, I'd get 
the crash -- again -- and be able to give you "similar" data from a 
technically different crash.

But it didn't crash.

this is with the SAME active cygwin1.dll as all those other crashes.  I 
then figured it must be due to some wierd way make, gcc, and cygwin 
interact with given '-g' instead of '-O2'.  So, I packaged up my nice 
new "-g newlib"-based kernel stuff into tarballs, intending to install 
it later and then re-run these tests -- on a new crash under the 
-g(newlib) kernel -- at some point in the future.

But first, I wanted to re-rebuild cygwin, with the same active 
cgywin1.dll that had been running during all those previous crashes, 
using the same configuration settings that I had used in all those 
previous crashes.  Call it a sanity check.

Well, it's official: I am now totally insane.  It didn't crash.  I tried 
it again.  It still didn't crash.  I just don't get it.

I now have three pending cygwin1.dll's to test --
   1) reproduce the crash with my current "original" 20010902 cygwin 
kernel.  This had been crashing consistently when building kernels for 
the last few days.  *Where* in the build it crashed was different each 
time, but it always crashed.

   2) reproduce the crash with my "-g(newlib)" version of the 20010902 
kernel (I know, the kernel is *normally* compiled with -g; but newlib 
isn't.  By "-g kernel" I really mean "kernel where both the winsup code 
AND newlib BOTH were compiled with -g")

   3) Chris sent me a cygwin1-cgf.dll offline, and I need to reproduce 
(or hopefully NOT) on that kernel.

But I can't even do #1 -- which had been SEGV'ing repeatedly for days. 
Until just now.  I can't figure out what's different -- all I did was 
clean out the build tree (and I've done that many times before), rebuild 
with "-g" -- but NOT install the resulting kernel, and clean out the 
build tree again.  And now, I can't get it to crash.

AAAaaaggghhhh!!  I'm going to bed.

--Chuck






Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]