Issue647 and 648 - FreeBCP

[mmcnabb-vms]

The commits here are:

• First two commits: add environment variable to suppress the CRT debugging spam (there's various inconsequential memory leaks by default in freebcp) - annoying to scroll past those every time you're using freebcp
• Issue 648: the call to bcp_init() actually performs the empty-rowset query and sets up bcp_columns() already, it's redundant for freebcp.c to go and do that again. Removing that redundant code actually fixes issue 647 as well (if -E is not used)
• Issue 647: fix Native format if -E is used. The underlying cause is that commit 9058123 added logic to dbresults() to free tds->res_info on completion, so subsequent calls to dbcoltype() fail. I avoided the problem by performing the bcp_colfmt before making the queries for the -E switch.
• Two refactoring commits to make the code more readable . I noticed while doing this that the BCPBATCH command was set for Character files but not for Native or Format files, not sure what the rationale was for that.

I have done various ad-hoc testing with SQL Server in different operating systems with large tables with varied column types, and not found any problems.

[mmcnabb-vms]

Update: have got my ASE 16.0 SP04 evaluation working again; and using the Format file appears to work correctly, although I only did an ad-hoc test with a simple table containing an int and a datetime.

[freddy77]

Please not, FreeBCP should use dblib, not abuse it.

[mmcnabb-vms]

OK, have copied what was done for bcp_getbatchsize to avoid this include. Pushed fix on top of the commit you commented on, but before my refactor.

I guess for future maintainability we may want to consider changing these two functions into something like bcp_control but for reading instead of writing.

[freddy77]

First 2 commits merged

[mmcnabb-vms]

There's been a fair bit of progress on "master" since I opened this, I'll squash the dblib.h fix and rebase . Also fixed problem with mixed code/declarations that failed the VS2008/2010 appveyor