Created attachment 334035 [details]
Much abbreviated .gnucash with example commodity and account definitions

Hi,

First, many thanks to all who created and maintain Gnucash - **much** appreciated.

I am several months into a project to convert 10+ years of history from MSMoney into Gnucash, and have reached the stage where I am importing (relatively) recent investment transactions via OFX. I am encountering a problem with the wrong commodity being "matched" by the import matcher.  As a consequence Gnucash can't find the corresponding stock account, and so asks me to nominate or create an account.  If I select the correct stock account, it is rejected (because it has a different commodity! aka the correct commodity not the wrongly matched one) and the only "solution" from within the Gnucash gui I've found is to create a new stock account, then clean everything up later.  Painful and error prone since it can happen to several different stocks in a single import, and I still have several years worth of imports to do.

I do however *think* I have tracked down the line of C code causing the problem with the help of the debug trace files.  (I say think because I haven't yet managed to compile Gnucash from source on Windows yet, despite several days of reading and following the Gnucash-on-Windows instructions as closely as I can - I will get back to that after posting this report - in the meantime, I am hoping that an "official" patch and compiled version? might become available).

The scenario I am encountering has been reported before, see Bug - 648523 - "OFX import account match failure", but as far as I can ascertain, it is still open.

In more detail.
==============
My gnucash file holds two commodities, one with CUSIP AGK and one with CUSIP AGKRA.  There are two different STOCK accounts, corresponding to each of these commodities. See attached much abbreviated .gnucash file).  

I have an OFX investment transaction (a SELLSTOCK) that pertains to the shorter of these two CUSIPS (namely to AGK). (I will attach the OFX file when I figure out how supply more than one attachment).  

The trace file (again will attach as soon as I can) shows that this is being matched not to the AGK commodity, but to the that AGKRA commodity.

* 11:33:08 DEBUG <gnc.import> [gnc_file_ofx_import] Filename found: C:\Users\Geoff\Documents\MSMoney data files\MSMoney_Data_Programs\Downloads for OFX\Encircle\Encircle OFX\20160819_20160819_160720 TS 141001_141031_pdf2txt__statement_33001594.OFX
* 11:33:08 DEBUG <gnc.import> [gnc_file_ofx_import] Opening selected file
* 11:33:08 DEBUG <gnc.import> [gnc_import_select_commodity] Default fullname received: AGL ENERGY - ORDINARY - (AGK)
* 11:33:08 DEBUG <gnc.import> [gnc_import_select_commodity] Default mnemonic received: AGK
* 11:33:08 DEBUG <gnc.import> [gnc_import_select_commodity] Looking for commodity with exchange_code: AGK

....

* 11:33:08 DEBUG <gnc.import> [gnc_import_select_commodity] Looking at namespace ASX
* 11:33:08 DEBUG <gnc.import> [gnc_import_select_commodity] Looking at commodity ANZ TD 4.500 3.0 Mths - (ANZTD450030M)
* 11:33:08 DEBUG <gnc.import> [gnc_import_select_commodity] Looking at commodity CWLTH BANK - CN 3M PER Q RD T-24 - (CBAPD)
* 11:33:08 DEBUG <gnc.import> [gnc_import_select_commodity] Looking at commodity Anz Bank - Pref Share Cps1 - (ANZPB)
* 11:33:08 DEBUG <gnc.import> [gnc_import_select_commodity] Looking at commodity ANZ BANK - CTG PRE 3M PER T - (ANZPA)
* 11:33:08 DEBUG <gnc.import> [gnc_import_select_commodity] Looking at commodity Commonwealth Bank - Perp.ExcPERLS V - (CBAPA)
* 11:33:08 DEBUG <gnc.import> [gnc_import_select_commodity] Looking at commodity Westpac Coupon Select Deposit - (WCSD)
* 11:33:08 DEBUG <gnc.import> [gnc_import_select_commodity] Looking at commodity Austrust cash deposit fund (Tower) - ()
* 11:33:08 DEBUG <gnc.import> [gnc_import_select_commodity] Looking at commodity RIO TINTO - ORDINARY - (RIO)
* 11:33:08 DEBUG <gnc.import> [gnc_import_select_commodity] Looking at commodity Westfield Grp Stapled Securities - (WDC)
* 11:33:08 DEBUG <gnc.import> [gnc_import_select_commodity] Looking at commodity Westfield America $3 Cap Notes - ()
* 11:33:08 DEBUG <gnc.import> [gnc_import_select_commodity] Looking at commodity Westfield Holding Ltd Ord Fully Paid - (WSF)
* 11:33:08 DEBUG <gnc.import> [gnc_import_select_commodity] Looking at commodity AGL ENERGY - RTS 08-SEP-14 FORUS - (AGKRA)
* 11:33:08 DEBUG <gnc.import> [gnc_import_select_commodity] Commodity AGL ENERGY - RTS 08-SEP-14 FORUS - (AGKRA) matches.

The remainder of the problematic behaviour arises from this incorrect match.

The match is done by code in src/import-export/import-commodity-matcher.c

The segment of code which reads:

            if (gnc_commodity_get_cusip(tmp_commodity) != NULL &&
                    cusip != NULL &&
                    strncmp(gnc_commodity_get_cusip(tmp_commodity), cusip, strlen(cusip)) == 0)
            {
                retval = tmp_commodity;
                DEBUG("Commodity %s%s", gnc_commodity_get_fullname(retval), " matches.");
            }

is what determines whether or not a match has been found.  In this code "cusip" is the string from the OFX file (in my example AGK), and (I presume) "gnc-commodity_get_cusip(tmp_commodity)" is the CUSIP of the candidate commodity in the current iteration of the search loop.  The bug occurs if this loop reaches the AGKRA commodity *before* it reaches the correct AGK commodity, because the strncmp test only looks at "strlen(cusip)" characters, ie only at 3 characters in my scenario.  If instead it looked at all the characters (ie at "max(strlen(cusip), strlen(gnc_commodity_get_cusip(tmp_commodity))" , the incorrect match would not occur. 

It seems this problem could occur whenever the cusip of one commodity is a leading substring of the cusip of another commodity - but apparently intermittently, depending on which one is encountered first in the search loop. I have not been able to determine the ordering of the internal table through which Gnucash is searching.

(PS From reading I believe that C does not have a "max" function, so please consider that last as pseudo-code).

I hope that explanation is clear (and enough to help a patch be developed).  I'm not very familiar at all with C but have tried to code a corrected version of import-commodity-matcher.c to check my reading of the code.  However, as mentioned earlier, I am still struggling to get the 'out-of-the-GitHub-box' Gnucash code to compile on my Windows machine, so as yet have not been able to try out my patched version.

Once again, thanks to all who look after Gnucash.

Regards

Geoff