BNC World: Reported Problems

This page lists all problems reported by users of the BNC World Edition since distribution began in February 2001. Wherever possible, a fix or an explanation is given. Please check this page before reporting any problems of your own! If you have any problem installing the software or the corpus, please report it to natcorp@oucs.ox.ac.uk

Problems installing the corpus

Some problems have been reported by people who have failed to complete the installation. Later versions of Windows sometimes obscure the DOS window in which the unpacking of the corpus files takes place, with the result that you may think it has finished before it has.

A complete installation of the corpus should contain folders A to K (but omitting I). The last file in K should be KS/KSW.

Problems installing SARA

SARA under Windows (single user licence)

A few people have reported problems installing the corpus caused by the fact that the process appears to have finished before it in fact has. If you are doing a complete installation of SARA (texts, plus indexes) do not remove the first CD from the drive until prompted to insert the second one, and do not assume that the installation is complete until all DOS windows are closed. Unfortunately, the installation procedure we used is not very clever in its use of DOS files...

If you are having problems with the CD installation procedure, you may like to try our alternative manual installation procedure

SARA under Unix (networked licence)

Warning messages appear in the compiler output

The messages about incompatible pointers in "QADLims" and "PrevWord" or "NextWord" may all safely be ignored. The message deprecating the use of the "gets" function may also safely be ignored.


The solve client distributed with the CD appears to give correct results even when the server is not running

The solve client distributed on the CDs attempts to connect to a SARA server running on a host called janus.oucs.ox.ac.uk instead of to your local system.

Circumvention: Edit line 32 in solve.c to read as follows:

#define BNC_HOST "localhost"
and recompile; or download this replacement source


The server cannot be built on a big endian machine using the supplied Makefile because of missing external references

Apologies: a defective version of the Makefile was shipped. Please download this replacement source for the big endian Makefile.


The server fails to start, reporting an incorrect or invalid signature file in the testlog.txt

This problem has been reported on a variety of servers built on big-endian machines: the most likely cause is that your system implements the network byte ordering routines in a different way from ours. The circumvention is to download this replacement source for the file endian.c, and also download this replacement source for the file corphash.c, and then rebuild the server.

Problems running the windows client

Scope restrictions for the <s> element do not work

If you change the scope node in a query builder to <s> in order to ensure that the rest of the query is satisfied within a single s-unit, the server always returns "no solutions". This is because the end-tags for these elements are still missing from BNC-world. This modified version of corpmain.c causes the server to recognize missing end-tags for <s> elements. Download it and rebuild the server to solve this problem.


Listings file saved in XML format contain erroneous characters

The ampersand character is incorrectly represented in XML format listings file. This is fixed in this new version of the client. (This new version contains a rebuilt server.dll which includes all other relevant fixes for standalone SARA usage)


Words with frequencies greater than 32,000 are not displayed bywordquery

When the word query box is used to select words by frequency (by checking any of the options on the Download tab other than the first), high frequency words are invisible. This is because some of the internal files used by the server are incorrectly sorted. The problem is only apparent when selecting words by frequency: it can be fixed either by rebuilding the indexer and regenerating the files concerned, or by replacing the files affected. For more information read this patch file.


Registering a subcorpus is impossible without use of the indexer program, which is not included in the distribution

The MS-DOS version of the indexer program is now included with the current version of the client.


Last changed: 23 July 2001