Ok, I got a little bit further. I'm able to get it to spit out at least some more information:
Z:\Users\hww3\Pike-v7.6.86\build\windows_xp-5.1.2600-i686>pike -t10 - DNOT_INSTALLED -DPRECOMPILED_SEARCH_MORE -m "c:\Program Files\pike\lib \master.pike" - -: program(Users\hww3\Pike-v7.6.86\src\modules \files\udp.c:986)() - -: returns: object(Users\hww3\Pike-v7.6.86\src \modules\files\udp.c:986) - -: program(Users\hww3\Pike-v7.6.86\src\modules \files\udp.c:1010)() - -: returns: object(Users\hww3\Pike-v7.6.86\src \modules\files\udp.c:1010) - -: 15483d4->__INIT() \Users\hww3\Pike-v7.6.86\src\port.c:699: Fatal error: Buffer overflow in VSPRINTF. Got exception when trying to describe backtrace.
Turns out, the problem is that on my build machine (OSX/Tiger Intel), sed doesn't like the magic cookie used in master.pike.in, complaining about illegal byte sequences. As a result, master.pike was an empty file, causing Pike.exe to bail out. Any chance we can trap that problem (perhaps printing an error to that effect) rather than segfaulting?
I've run verify, and it fails 2 tests:
/Users/hww3/Pike-v7.6.86/src/testsuite.in:9212: Test 6889 (shift 2) (CRNL) failed. 1: mixed a() { foreach (({1075550400,94691300,220921700,347152100,473382500, 2: ^I^I 599612900,725843300,852073700,978304100,1104534500, 3: ^I^I 1230764900,1356995300,1483225700,1609456100,1735686500, 4: ^I^I 1861916900,1988147300,2114377700 5: ^I^I }),int t) if(mktime(gmtime(t))!=t) return t; return 0;; } 6: mixed b() { return 0; } 7:
Error: Time conversion failed. Unknown program: mktime(20,8,23,31,11,136,0,0) testsuite:5: testsuite()->a() /Users/hww3/Pike-v7.6.86/bin/test_pike.pike:776: /Users/hww3/Pike- v7.6.86/bin/test_pike()->main(3,({"/Users/hww3/Pike-v7.6.86/bin/ test_pike.pike",0,"testsuite"}))
Testing directory "Z:/Users/hww3/Pike-v7.6.86/build/ windows_xp-5.1.2600-i686/test-install/pike/7.6.86/lib/modules"...
the second failed test looks like missing symbols in the COM module:
DL: Symbol '_GUID_NULL' not found. DL: Symbol '_wcstombs' not found. DL: Symbol '_string_builder_vsprintf' not found. DL: Symbol '_IID_IDispatch' not found. DL: Symbol '_GUID_NULL' not found. DL: Symbol '_GUID_NULL' not found. DL: Symbol '_GUID_NULL' not found. DL: Symbol '_string_builder_vsprintf' not found. DL: Symbol '_IID_IDispatch' not found. DL: Symbol '_IS_ZERO' not found. DL: Symbol '_IS_ZERO' not found. DL: Symbol '_IS_ZERO' not found. DL: Symbol '_IID_IDispatch' not found. DL: Symbol '_IID_IDispatch' not found. DL: Symbol '_IID_IPersistFile' not found. DL: Symbol '_IID_IDispatch' not found. test: failed to load "Z:/Users/hww3/Pike-v7.6.86/build/ windows_xp-5.1.2600-i686/test-install/pike/7.6.86/lib/modules/ COM.so": load_module("Z:/Users/hww3/Pike-v7.6.86/build/ windows_xp-5.1.2600-i686/test-install/pike/7.6.86/lib/modules/ COM.so") failed: Symbol '_IID_IDispatch' not found.
Bill
In the last episode (Jul 17), H. William Welliver III said:
Turns out, the problem is that on my build machine (OSX/Tiger Intel), sed doesn't like the magic cookie used in master.pike.in, complaining about illegal byte sequences. As a result, master.pike was an empty
If you set LANG=C does the sed command work?
Nope. I had to install GNU sed. No biggie, but i suspect that it's the FreeBSD sed, so it might be good to be aware of.
Bill
On Jul 18, 2006, at 12:51 AM, Dan Nelson wrote:
In the last episode (Jul 17), H. William Welliver III said:
Turns out, the problem is that on my build machine (OSX/Tiger Intel), sed doesn't like the magic cookie used in master.pike.in, complaining about illegal byte sequences. As a result, master.pike was an empty
If you set LANG=C does the sed command work?
-- Dan Nelson dnelson@allantgroup.com
I'm having problems with the COM module as well. We probably need somthing in port.h for _GUID_NULL and _IS_ZERO. Or we just make a configure test that checks for those and disables the module.
Until somebody finds a "string_builder_vsprintf" in the Pike source code, my bet is on this configure test being appropriate:
if true; then PIKE_FEATURE_RAW([COM],[no (broken)]) pike_cv_com = no fi
pike-devel@lists.lysator.liu.se