I have found the cause of the problem at the BEN site : The EMBL has 3
sequences of more than 350000 bases long and from GCG 8.1 on, the way
GCG splits these sequences has changed. To allow SRS to recompose the
sequences, I modified the file srsgeneral.sdl, I replaced
! gcg sequence file
#filetype /ID=%GCGSEQ_FILE
/typename=seq /maxline=20000
/begstr=">>>>" /advance=0 /exitstr=">>>>"
/find=id /iscont="-"
by
! gcg sequence file
#filetype /ID=%GCGSEQ_FILE
/typename=seq /maxline=20000
/begstr=">>>>" /advance=0 /exitstr=">>>>"
/seqtype=gcg8_1
I had not reindexed. Changing srsgeneral.sdl back to its old form
cures the problem of the faulty sequence retrieval, except of course
that the long sequences are not handled well.
I do not know if remaking all the indexes after the change would have
helped (since this takes a whole weekend, I did not try it just now).
I also have no idea if the problem of Christoph Gartmann has the
same cause or is entirely different.
Guy Bottu