[ILUG] Opera8 loop/zoom bugs -vs- Linux and "reverse engineering"?

[Brian Foster]

 I normally use Opera as my GUI web browser,
 and recently downloaded the free version of
 the newly released version 8 to two different
 systems:  SUSE 9.1 (kernel 2.6), and an ancient
 Debian (kernel about 2.4.17(?)).  on the Debian
 box, the Opera8 was a clean install; i.e., that
 box had never had Opera installed before.

 two problems have become quite apparent.  both
 problems are seen on both boxes:

 ①  on some sites — the Irish Times ireland.com
   seems to be ‶good” for this — zooming on some
   pages causes Opera8 to segfault. ;-(  a search
   of the forum pages on the Opera site suggests
   that at least one other person has the same
   problem (that person is also running Linux).
   I have no investigated this problem further.

 ②  even when “idling” (i.e., not being used,
   no (known) JS running, not connected, &tc),
   one of the threads in Opera8 is executing
   a very tight loop doing gettimeofday(2)s
   with an occasional write(2), ioctl(2), or
   time(2) at not-obvious intervals (the data
   following was all collected on the SUSE box):

$ strace -ttt -T -p ...  |  egrep -C 2 -v gettimeofday
 ...
1115023795.813194 gettimeofday({1115023795, 813233}, NULL) = 0 <0.000016>
1115023795.813295 gettimeofday({1115023795, 813319}, NULL) = 0 <0.000015>
1115023795.813371 write(8, "\0", 1) = 1 <0.000021>
1115023795.813498 gettimeofday({1115023795, 813520}, NULL) = 0 <0.000015>
1115023795.813564 gettimeofday({1115023795, 813584}, NULL) = 0 <0.000015>
--
1115023796.314113 gettimeofday({1115023796, 314146}, NULL) = 0 <0.000027>
1115023796.314196 gettimeofday({1115023796, 314217}, NULL) = 0 <0.000015>
1115023796.314394 ioctl(5, FIONREAD, [0]) = 0 <0.000023>
1115023796.314529 write(6, "\0", 1) = 1 <0.000018>
1115023796.314617 gettimeofday({1115023796, 314639}, NULL) = 0 <0.000015>
1115023796.314686 gettimeofday({1115023796, 314707}, NULL) = 0 <0.000015>
--
1115023798.239510 gettimeofday({1115023798, 239532}, NULL) = 0 <0.000015>
1115023798.239579 gettimeofday({1115023798, 239600}, NULL) = 0 <0.000015>
1115023798.239666 time([1115023798]) = 1115023798 <0.000013>
1115023798.239729 time([1115023798]) = 1115023798 <0.000012>
 ...[ c.10 calls to time(2) deleted ]...
1115023798.241189 time([1115023798]) = 1115023798 <0.000011>
1115023798.241245 time([1115023798]) = 1115023798 <0.000012>
1115023798.241302 gettimeofday({1115023798, 241323}, NULL) = 0 <0.000015>
1115023798.241374 gettimeofday({1115023798, 241395}, NULL) = 0 <0.000016>
 ...

   gdb(1) indicates the defective Thread is using
   QEventLoop::processEvents():

(gdb) where
#0  0xffffe410 in ?? ()
#1  0xbfffe958 in ?? ()
#2  0x088b5074 in ?? ()
#3  0x088b4ef0 in ?? ()
#4  0x4110a391 in ___newselect_nocancel () from /lib/tls/libc.so.6
#5  0x4093dab7 in QEventLoop::processEvents () from /usr/lib/libqt-mt.so.3
#6  0x409a6811 in QEventLoop::enterLoop () from /usr/lib/libqt-mt.so.3
#7  0x409a6656 in QEventLoop::exec () from /usr/lib/libqt-mt.so.3
#8  0x409906cf in QApplication::exec () from /usr/lib/libqt-mt.so.3
#9  0x08420bc5 in jzero_far ()
#10 0x08063039 in ?? ()
 ...
#62 0xbfffedf8 in ?? ()
#63 0x40ca864c in QGDict::look_ascii () from /usr/lib/libqt-mt.so.3
#64 0x4106e500 in __libc_start_main () from /lib/tls/libc.so.6
#65 0x08061ead in ?? ()
(gdb) 

 has anyone else  ⒜ observed this behaviour?
                  ⒝ can confirm it?
                  ⒞ have suggestions for what other details
                     to include in the defect report to Opera?
                     (besides the obvious like versions et.al.)
                  ⒟ have any idea w.t.f. is going on???
 the 2nd problem can be clumsily worked-around by
 `kill -STOP'ing Opera8 when not-in-use; but the
 1st has no known/easy workaround (except zooming _before_
 you open the URL, which is obviously not too practical).

 also, the Opera license prohibits disassembly and reverse-
 engineering:

    “Software means Opera, all program and information files and
     other documentation which are part of the Opera Software
     package [ ...]

    “You shall not modify, translate, reverse engineer, decompile
     or disassemble the Software or any part thereof or otherwise
     attempt to derive source code or create derivative works
     therefrom.”

 I do not believe anything I've done above runs afoul of that
 (e.g., /usr/lib/libqt-mt.so.3 is not distributed as part of
 Opera8 software package (for SUSE 9.x at least)), but... how
 far can I go?  at what point does this sort of informal defect
 investigation turn into activities that could be reasonably
 construed as violating such license restrictions?

 I am aware Opera is not FOSS.  in this case, it does not
 matter (to me), at least for problem 2, as I do not have
 GUI skills, and it smells like a GUI problem.  so please
 do not harangue about using non-FOSS software!

cheers!
	-blf-
-- 
Experienced (20+ yrs) kernel/software Eng: | Brian Foster   Montpellier,
 • Unix, embedded, &tc;  • Linux;  • doc;  | blf at utvinternet.ie   FRANCE
 • IDL, automated testing, process, &tc.   |  Stop E$$o (ExxonMobile)!
Résumé (CV) http://www.blf.utvinternet.ie  |     http://www.stopesso.com