Pike-devel

pike-devel@lists.lysator.liu.se

1 participants
3045 discussions

Re: new ground with windows
by Martin Stjernholm, Roxen IS ＠ Pike developers forum 23 Jul '06

23 Jul '06

You have to use -MD, both for pike.exe and all modules, when you use dlls. That's the only way to make them share the same CRT instance, which is vital for memory management to work. I suggest you try with --enable-dll (assuming a recent 7.7) which fixes that and other things. Note that I'm still not finished with the dll stuff. Notably I have a lot of uncommitted changes to make the installer work. Haven't done much about it lately since I'm on vacation and mostly off-net (will still be for another week).

1 0

new ground with windows
by H. William Welliver III 23 Jul '06

23 Jul '06

Ok, I've managed to compile one of my C modules on windows (the XML2 parser glue), but I'm running into problems with dynamic module loading in pike. The libxml2 module uses some functions like modf and log10. Because libxml2 is being compiled as a library, the c library doesn't appear to be linked. Similarly, when I make the module, these symbols (and the snprintf pointer in the following error message) don't seem to be getting linked in. Thus, when I try to load the compiled module into Pike, I end up with undefined symbols. The question is, how do I work around it? I can get the regular functions like log10 and fmod to be linked into Pike by creating dummy functions that use them, but then that will only work with my copy of pike and not anyone elses. Similarly, I'm not sure what to do with the __imp records. Is dlopen not aware of how to resolve these? Any thoughts and suggestions would be greatly appreciated! Bill DL: Symbol '__imp___snprintf' not found. DL: Symbol '__imp___snprintf' not found. DL: Symbol '__imp___snprintf' not found. DL: Symbol '__imp___snprintf' not found. DL: Symbol '__imp___snprintf' not found. DL: Symbol '__imp___snprintf' not found. DL_Symbol '_modf' not found. DL: Symbol '_log10' not found. DL: Symbol '__imp___snprintf' not found.

2 1

-mcpu
by Martin Nilsson (Opera Mini - AFK!) ＠ Pike (-) developers forum 23 Jul '06

23 Jul '06

I get `-mcpu=' is deprecated. Use `-mtune=' or '-march=' instead. for every file comiled in MinGW. Is there anything someone more configure proficient can do to reduce the spam?

2 1

/usr/local/include
by Martin Nilsson (Opera Mini - AFK!) ＠ Pike (-) developers forum 23 Jul '06

23 Jul '06

While trying to compile Pike under MinGW I've encountered the problem that include files in /usr/local/include isn't found. First place of error is in the Gmp module. configure:1574: checking for gmp.h configure:1584: gcc -E conftest.c >/dev/null 2>conftest.out configure:1580:17: gmp.h: No such file or directory configure: failed program was: #line 1579 "configure" #include "confdefs.h" #include <gmp.h> If I manually add -I/usr/local/include it works, but what is the correct, general solution?

3 5

SVN
by Marcus Comstedt (ACROSS) (Hail Ilpalazzo!) ＠ Pike (-) developers forum 23 Jul '06

23 Jul '06

So, how about that switch to Subversion, eh? I'd very much like to rename lib/modules/_Image.pmod to Image.pmod and src/modules/Image to _Image, to get rid of image_magic_index, and with svn I could do that without messing everyones working copy up...

10 19

further debugging of pike/windows
by H. William Welliver III 19 Jul '06

19 Jul '06

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

4 4

nettle lib for windows
by Bill Welliver 14 Jul '06

14 Jul '06

anyone have a pre-compiled version of nettle (or a know working copy of gmp) for windows? I managed to get the remote compilation process working, but neither of the bundles want to compile properly. Bill

3 7

Re: SVN
by Johan Sundstr�m (Achtung Liebe!) ＠ Pike (-) developers forum 10 Jul '06

10 Jul '06

Then I understand. It fits your concept of pragmatism, while not mine nor Peter's, and it ticks off Martin Bähr's ideas of freedom. :-) All that said, and without myself having dived into technicalities in a FishEye installation, I'm sure people would use the system if you brought it online somewhere and it wasn't required to run off the not-quite-state-of-the-art pike site. (Subversion has good APIs that make most things nonnecessary to run on the repository machine, so I'm assuming it at least possible. Though if it were, we would probably not have this thread in the first place.) The reasoning to use any tool short of perfection is that it scratches some itch better than not using a tool at all, and those imperfections are addressable with sa[nf]e licenses. Trac would have been great and already set up here, had it been comfy in a solaris locale or had we had operations linux based. FishEye may be neat if somebody sets it up somewhere and it just solves issues as well. We're no worse meritocracy than any other hacking camp, so "not invented here" never matters to "not maintained here" software. What qualifies as comfy rarely changes with arguments, though, so trying to convince any voice in the thread to embrace FishEye is probably moot.

3 3

Re: SVN
by Peter Bortas ＠ Pike developers forum 10 Jul '06

10 Jul '06

>> 3&4 How will evaluating it give me power to change the code when in >> fact the NDA forbids med to change the code? > >How will being able to change the code help you to evaluate the >software's feature set? I just don't see why you'd need to have that >kind of access. There are lots of other projects that seem to be >happily using the tool, so I'm just wondering why this project is >different. I have already evaluated the feature set. Twice. It's mostly good from what I can tell from the webdemos. It lacks a few things _that we can not add_. >I've used Trac as well, and while it's probably better than the >current situation, I think it tries to do too many things, and ends >up not being as useful. If there's a desire to solve multiple >problems (source browsing, bug ticketing and collaboration, I'd >recommend doing Fisheye + Mantis + some other Wiki tool, such as >FinScribe. There is no such wish that I know of even if it would be neat. It just happens that Trac beats out the other Open Source solutions as a reposotory browser and changeset viewer. >If I remember correctly, someone set up Mantis for our use >a year or two ago, but a decision was made to not (or no decision was >made to) use it in favor of Crunch (which I find to be hard to locate >easily and kinda difficult to use). There was a test installation but no decision was made to use it one way or the other. It was apparent at the time that noone had the time or interest to see if we could really use it. I found it to be very nice on first impression. >I'm not trying to make personal attacks, that was definitely not the >point. I've been involved with enough product evaluations to know >that arguments of the type you listed have resulted in plenty of >worthy competitors being disqualified when they might not need to >have been. And I have been involved in enough product evaluations to know that most packages are crap in one way or the other and deserved to be disqualified. You just have to decide what areas you don't want it to stink at and find the one that looks like the least amount of pain.

1 0

/usr/local/include
by Marcus Comstedt (ACROSS) (Hail Ilpalazzo!) ＠ Pike (-) developers forum 09 Jul '06

09 Jul '06

This is what appears in my config.log: [...] configure:56206: checking /usr/local/include configure:56249: gcc -E -I/home/marcus/Pike/7.7/build/linux-2.6.8dma-ppc/bundles/include -I/usr/local/include conftest.c configure:56255: $? = 0 configure:56268: result: added [...]

1 0

Jump to page:

2026

2025

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

2007

2006

2005

2004

2003

2002

Pike-devel