[Click] [click]: many tests don't work anymore...

[Eddie Kohler]

Whoops!  That problem is fixed.  Thank you for the bug report.

Eddie


Lionel Debroux wrote:
> Hello,
> 
> Today, I've cloned the git repository of Click. Then, I've built Click with the procedure
> described in INSTALL (common to all programs that use autotools):
> ./configure (no options)
> cd userlevel
> make
> 
> I'm running SimplyMEPIS 7.0 (based on Debian stable) with patched GCC 4.1.1 on a Core 2 Duo @ 2
> GHz with 2 GB of RAM.
> Next, I've of course tried to test it:
> ~/click/userlevel$ ./click ../conf/test.click
> Segmentation fault
> 
> Huh ?! Have I miscompiled the program ?
> 
> 
> I reconfigured and recompiled with -g3 for maximum debug info, and ran GDB. The backtrace was as
> follows:
> (gdb) run ../conf/test.click
> Starting program: /home/Lionel_Debroux/click/userlevel/click ../conf/test.click
> 
> Program received signal SIGSEGV, Segmentation fault.
> 0x08276211 in configure_order_compar (athunk=0x8351ad8, bthunk=0x8351adc, copthunk=0x0) at
> ../lib/router.cc:832
> 832         const int* configure_order_phase = (const int*) copthunk;
> (gdb) bt
> #0  0x08276211 in configure_order_compar (athunk=0x8351ad8, bthunk=0x8351adc, copthunk=0x0) at
> ../lib/router.cc:832
> #1  0x082634fd in click_qsort (base=0x8351ad8, n=6, size=4, compar=0x8276200
> <configure_order_compar>, thunk=0x8351af8)
>     at ../lib/glue.cc:468
> #2  0x08280479 in Router::initialize (this=0x8354c78, errh=0x8346268) at ../lib/router.cc:890
> #3  0x08253061 in parse_configuration (text=<value optimized out>, text_is_expr=<value optimized
> out>, hotswap=false, errh=0x8346268)
>     at click.cc:285
> #4  0x08253715 in main (argc=Cannot access memory at address 0x0
> ) at click.cc:440
> 
> 
> Since lines 832-833 of lib/router.cc are:
> const int* configure_order_phase = (const int*) copthunk;
> return configure_order_phase[*a] - configure_order_phase[*b];
> and copthunk = NULL, the program segfaults, as it should.
> 
> Further tests show that conf/test-device.click segfaults too.
> 
> 
> Before making a bisect, I browsed through the recent revision history:
> * as of now, the HEAD of the git repository is commit e9bb2123d5dfa6f14076a20c37fc501737a819aa
> "64-bit platforms need this".
> * a few lines below, I can see commit 75e13ed00c124cbde4054681d21f13d920628403 "Improve
> click_qsort implementation greatly.", followed by commit 
> 7e2e5b2b208ff4b137613adabaad5daad23df354 "Document the qsort change a touch more.". Both commits
> were made yesterday.
> 
> So I tried reverting those commits and recompiling. After that, conf/test.click and
> conf/test-device.click no longer crash.
> 
> I haven't tried yet to see what happens, but there's definitely something wrong with commit
> 75e13ed00c124cbde4054681d21f13d920628403.
> 
> Hope that information helps fixing the problem :)
> 
> 
> Regards,
> Lionel Debroux.
> 
> 
>       ____________________________________________________________________________________________
> Yahoo! Mail innove : interface hyper pratique, messenger intégré, couleurs - http://mail.yahoo.fr