I thought it should be possible to discover the used read structure and barcodes.

You can take a look at RunInfo.xml file in the top level directory which shows you the read structure.

                 <Reads>
                        <Read Number="1" NumCycles="150" IsIndexedRead="N"/>
                        <Read Number="2" NumCycles="8" IsIndexedRead="Y"/>
                        <Read Number="3" NumCycles="8" IsIndexedRead="Y"/>
                        <Read Number="4" NumCycles="150" IsIndexedRead="N"/>
                </Reads>

As for the barcodes you can simply call bases using a blank samplesheet to generate "Undetermined" read files with bcl2fastq. Then use the code here to identify indexes present in your data: C: Demultiplexing reads with index present in the labels

After that you can either go back and make up a samplesheet for a re-run with bcl2fastq or use demuxbyname.sh from BBMap suite to demultiplex your data. All this works as long as you have information about Sample_ID <--> barcode information.

Just convert the entire sequences to fastq without de-multiplexing and without trimming.

Note that you need to set the entire read length to template.

148T in this example.

picard  IlluminaBasecallsToFastq B=./{MY_RUN}/Data/Intensities/BaseCalls/ L=1 RS=148T INCLUDE_NON_PF_READS=false COMPRESS_OUTPUTS=true RUN_BARCODE=MY_RUN OUTPUT_PREFIX=MY_RUN READ_NAME_FORMAT=ILLUMINA  NUM_PROCESSORS=1 IGNORE_UNEXPECTED_BARCODES=false FORCE_GC=false

Then grep on the expected barcodes to see where they are located.