Non-standard header tag values are no longer typed.  They are all stored as strings.  Any existing files that have a non-standard header tag value with a value type encoded in it will need to be repaired.  These header tags will parse, but may not yield the expected results.  

AbstractSAMHeaderRecord.setAttribute(String, Object) has been deprecated.  Use AbstractSAMHeaderRecord(String, String) instead.

AbstractSAMHeaderRecord.getAttribute and .getAttributes have been changed to return String and Map<String, String>, respectively.  Existing code that called these methods may need to be changed, in the cases where other than a String return value was expected.

When merging the input files, if they are all sorted then use that to produce a sorted input iterator.

Emit warning if MD5 or BAM index creation is requested, but can't be done because output BAM is not a regular file.

Fix auto-MD5 generation, which wasn't working because of check that output file already exists rather than that it is a normal path.

CommandLinePrograms no longer display common options unless -H or --stdhelp is used or there was an error with a standard option.  Modified option and the CreateHTML source files to add a common attribute to identify options as common.

Tests changed to reflect FastqToSam no longer fails on read if trailing /1 or /2 is missing IF the read names are identical AND don't end in /1 or /2.  Also changed test execution process to include argument parsing in CommandLineProgram.

FastqToSam no longer fails on read if trailing /1 or /2 is missing IF the read names are identical AND don't end in /1 or /2

SamToFastq may now output fastq files per read group if OUTPUT_PER_RG (OPRG) is specified.  If OUTPUT_DIR is specified with OUTPUT_PER_RG then a file (two if paired end) will be output in OUTPUT_DIR per read group named after the platform unit for the read group.

When trying to find a sequence dictionary corresponding to a reference fasta, try filename.dict if basename.dict isn't found.  Patch from Nils Homer.