Re: [HACKERS] Access violation from palloc, Visual Studio 2005, C-language function
Kevin Flanagan kevi...@linkprior.com writes: Hard to tell without seeing the actual code and a stack trace, but I'd bet that you haven't fully resolved the build process problems you mentioned earlier. I've attached a zip of the (tiny) project, and a text file with the contents of the module containing the C-language functions. The only difference from sample code is that (as pointed out by Takahiro Itagaki in his post here of 8th March) the function implementations need decorating with __declspec(dllexport). Mph. I don't actually believe that, nor do I believe the #define BUILDING_DLL you put in, because neither of those are needed in any of our contrib modules. What I suspect at this point is that the reference to CurrentMemoryContext in the palloc() macro is being bollixed by having the wrong value for BUILDING_DLL. However, not having a Windows build environment to experiment with, I'll have to defer to somebody with more experience in that. (I wonder BTW if we should rename BUILDING_DLL, because it seems a bit misnamed. AIUI it's supposed to be set while building the core backend, not while building loadable modules.) regards, tom lane -- Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-hackers
Re: [HACKERS] Access violation from palloc, Visual Studio 2005, C-language function
Aha. I'd read that the build process for the contrib modules involved generating a .DEF file for the necessary exports. I had the impression that defining BUILDING_DLL was an alternative, addressing (part) of the issue (that is, PG_FUNCTION_INFO_V1 declares functions as 'extern PGDLLIMPORT', and if you define BUILDING_DLL, then PGDLLIMPORT is defined as ' __declspec (dllexport)'). But you're quite right, if I take out the BUILDING_DLL definition, and put the __declspec (dllexport) stuff in piecemeal, the access violation goes away. Thank goodness. Thanks, that really helped me out. Kevin. -Original Message- From: pgsql-hackers-ow...@postgresql.org [mailto:pgsql-hackers-ow...@postgresql.org] On Behalf Of Tom Lane Sent: 10 March 2010 18:51 To: Kevin Flanagan Cc: 'PostgreSQL-development' Subject: Re: [HACKERS] Access violation from palloc, Visual Studio 2005, C-language function Kevin Flanagan kevi...@linkprior.com writes: Hard to tell without seeing the actual code and a stack trace, but I'd bet that you haven't fully resolved the build process problems you mentioned earlier. I've attached a zip of the (tiny) project, and a text file with the contents of the module containing the C-language functions. The only difference from sample code is that (as pointed out by Takahiro Itagaki in his post here of 8th March) the function implementations need decorating with __declspec(dllexport). Mph. I don't actually believe that, nor do I believe the #define BUILDING_DLL you put in, because neither of those are needed in any of our contrib modules. What I suspect at this point is that the reference to CurrentMemoryContext in the palloc() macro is being bollixed by having the wrong value for BUILDING_DLL. However, not having a Windows build environment to experiment with, I'll have to defer to somebody with more experience in that. (I wonder BTW if we should rename BUILDING_DLL, because it seems a bit misnamed. AIUI it's supposed to be set while building the core backend, not while building loadable modules.) regards, tom lane -- Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-hackers -- Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-hackers
Re: [HACKERS] Access violation from palloc, Visual Studio 2005, C-language function
Kevin Flanagan kevi...@linkprior.com writes: Environment: Windows Vista, PostgreSQL 8.4 (1-click installer), Visual Studio 2005 sp1. I have a bare-bones DLL built as per the above, compiling the 'add_one' and 'copytext' samples found at http://www.postgresql.org/docs/8.4/interactive/xfunc-c.html (version 1 calling convention), compiled as 'C'. I can use 'add_one' just fine from within SQL, but if I use 'copytext', an access violation occurs as soon as palloc() is called. Could anyone suggest what the problem might be? Hard to tell without seeing the actual code and a stack trace, but I'd bet that you haven't fully resolved the build process problems you mentioned earlier. I'm thinking this may be a symptom of linkage failure, since palloc is probably the first place in the above-described sequence where your DLL is going to call back into the core backend. Another possibility is that you mistranscribed the example somehow. Maybe you forgot the PG_FUNCTION_INFO_V1(copytext) macro? Failing that, are there any other (creative?) ways to return strings from a C-language function without using palloc? If you can't make those examples work, you have fundamental problems you need to fix, not find a creative workaround. regards, tom lane -- Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-hackers