ogr-oci: ORA-00972: identifier is too long

[martinl]

OGR OCI driver doesn't check length of column names ​ORA-00972 when writing data. If the length is more than 30 characters than adding such column fails. The attached patch avoids such failure, see warning below:

Warning 1: Column BonitovanyDilBonitovanaJednotkaKod is too long (at most 30 characters). Using BonitovanyDilBonitovanaJednot_.

[martinl]

avoid ora-00972 for column names

[Even Rouault]

Martin,

apart szFieldName[sizeof( szFieldName )] = '\0'; which looks like a 1-byte-after-end-of-buffer and potential indentation issue (tabulation), I think you can go ahead with the patch.

[martinl]

Replying to rouault:

apart szFieldName[sizeof( szFieldName )] = '\0'; which looks like a 1-byte-after-end-of-buffer and potential indentation issue (tabulation), I think you can go ahead with the patch.

done in r27449. Both notes reflected. After some testing I will do backport (if no objections) to 1.11 branch.

Martin

[martinl]

Replying to martinl:

done in r27449. Both notes reflected. After some testing I will do backport (if no objections) to 1.11 branch.

backported in r27582. Closing this ticket.

[Even Rouault]

Martin, r27582 looks wrong at szFieldName[sizeof( szFieldName )] = '\0'; . It has the 1-byte-after-end-of-buffer write that r27449 as not. By the way, I think (should be verified, the snprintf man page isn't completely clear) that snprintf() always null terminate (contrary to strncpy !) so the line might be not needed at all.

[martinl]

Replying to rouault:

Martin, r27582 looks wrong at szFieldName[sizeof( szFieldName )] = '\0'; . It has the 1-byte-after-end-of-buffer write that r27449 as not. By the way, I think (should be verified, the snprintf man page isn't completely clear) that snprintf() always null terminate (contrary to strncpy !) so the line might be not needed at all.

ops, sorry, should be hopefully fixed in r27583.

[martinl]

Replying to martinl:

Replying to rouault:

Martin, r27582 looks wrong at szFieldName[sizeof( szFieldName )] = '\0'; . It has the 1-byte-after-end-of-buffer write that r27449 as not. By the way, I think (should be verified, the snprintf man page isn't completely clear) that snprintf() always null terminate (contrary to strncpy !) so the line might be not needed at all.

ops, sorry, should be hopefully fixed in r27583.

possible to close this ticket?