History of changes between different versions of the report generator
script dorep
Version 3.0
- Script name
The dorep script is named dorep3.pl now, because the version
has changed to 3.0 to draw level to the newest SKaMPI
version.
- New .dorep format
The format of the option files for dorep has changed. However,
the old format is still parsed correctly. The new format introduces
sections whose names have to be written in square brackets. Known
sections are:
- [Options]
This section contains definitions for dorep options as
before
- [Comparisons]
This section contains descriptions for suite comparisons as
before
- [Suites]
This section contains the suites which dorep shall
include in the non-comparison suite sections
- [Exclude]
This section contains the suites which dorep shall not
include in the non-comparison suite sections
Further details can be found in the README
file.
- Option -m more restrictive
To avoid gnuplot problems you must give a minimum data number of 2 or
more.
- New temporary .ski like files
In addition to the temporarily produced .ski files,
.ski.cut files are produced now. These only contain the
measurement lines of their respective .ski files. The
difference is that the node times are cut away from these lines. This
is due to a problem of gnuplot. This program has a problem
with lines of a certain length. When there are lots of node times,
gnuplot somehow messes around and produces REALLY BIG
.eps files (in fact it seems it would produce files of
infinite size, so you have to stop the execution of the program).
This dorep version just feeds gnuplot the
.cut files and everything should be alright.
- Defined exit values
If the script runs through without coming across an error, it exits
with the value 0. An error will leave the script with the value 1.
- LaTeX started in `non-stop mode' instead of `batch
mode'
In the previous version of dorep LaTeX was executed
in batch mode so that it wouldn't hang around waiting for input from
the user. LaTeX's non-stop mode behaves quite the same, but it
additionally puts the error messages on standard output, which I found
a little nicer. My goal was that the contents of the dorep.log
file were as informative as before dorep version 2.1.
- Changes in report output
- The possibility of defining units in which to measure is taken into
consideration now. Since these definitions can be quite long, they
are no longer printed inside the table which describes a suite's
variation parameters, but below this table. Another consequence of
non-standard units is that the message lengths and bandwidths are
no longer given in `bytes' but simply in `units' now.
- In SKaMPI you can define types based on basetypes. These
basetypes are--if used--described in a special section at the
beginning of the report.
- If no graphical output was possible for comparisons due to too few
data lines, former versions of dorep didn't produce
sections for those comparisons. Version 3.0 does this by producing
comparison tables similar to the ones used in non-comparison suite
sections where there were not enough data lines. Thus, the user can
also compare `simple' suites now.
- Wider barchart images will be produced if more than 63 node times
are shown.
- The title page has changed slightly. The date of the
SKaMPI run now is shown together with the other run
information. Below, the dorep version is printed together
with the date of the dorep run. (So now there's one
LaTeX warning less because of using an \author command
;-)
- Band-width is now bandwidth. (I hope it's correct now; I looked it
up and found it written tohether or separated--but not with a
hyphen) ;-)
- The default values output now includes the maximum/actual
times.
- Changes in console output
- The deleting of the temporary files can last quite some time, so
the user now is told that deleting is in progress.
Version 2.1
- Temporary files overwritten by default
CAUTION! Usage is different:
Existing temporary .tmd and .gpl files now are
overwritten by default. So the options -o and
--overwrite are no longer accepted. Instead, there are
options -u and --use_temp which tell the script
to use .tmd and .gpl files if existing.
- Barchart options possible at command line
The options -b and --barchart now are also
available at the command line. The parameters for these options have to
be one single parameter at the command line so that you have to enclose
them in quotes.
- No temporary files if institution is missing
The existence of an entry for the institution is checked earlier now so
that the script can be stopped before creating any temporary
files.
- No (or less) irritating warnings and errors
- The request for filling in some institution was misleading. The
user was told to do this in the .dorep file or similar,
which, of course, is total nonsense. This has to be done in the
skampi.out file, which the user is told correctly
now.
- Gnuplot gave a warning if all node times in a barchart
graphic were equal. The script is improved so that there should be
no more "empty y range" warnings.
- If anyone ever tried to supply a skampi.out file no
residing in the current directory, dvips would have failed
because it looked for the DVI file inside the directory of
the skampi.out file where it wasn't. Now dvips is
given the correct location of the file.
- Special characters in description strings in skampi.out
files now are prepared so that LaTeX shouldn't produce any
errors any more. (Which was especially nasty because LaTeX
output is written into a file so that you couldn't tell if the
script was still running or not.)
- By the way: LaTeX is executed in batch mode (first line in
LaTeX output file) so that it can't get stuck waiting for
the user to provide some error handling. It just runs through and
puts all error messages in its log file (skampi.out.log or
similar) where you can study errors. (Which of course shouldn't be
necessary ;-)
- The current version looks for dri-files at the beginning.
So if there aren't any files ending with .dri in the
respective directory, you get a warning once and aren't showered
with warnings of single missing files. Nevertheless, you get single
warnings if some dri-file is found, even if it isn't used
for the skampi.out file or it isn't even a dorep
information file.
- All produced files go into current directory
All the files - if temporary or not - go into the directory from which
the dorep script is called. In the earlier version the
LaTeX file was written into the same directory which the
skampi.out file resided in. This has something to do with the
dvips problem explained above.
- Possibility of deleting temporary files pointed
out
If the script ends without deleting the temporary files, a script
kill_temp is created. Executing this script deletes the
temporary files created by the latest run of dorep2.pl. If
kill_temp was created, dorep2.pl now tells the user
about this possibilty.
- Miscellaneous
Some things have been done to improve error tolerance or to improve
output:
- Empty entries for USER or other things in a
skampi.out file are now printed in a somewhat "more
correct" way. In fact, there should be no empty entries in the
skampi.out files!
- If all node times of a measurement are equal to 0 or even less, no
barchart graphics are produced for this measurement. The output of
node times less than 0 would surely be a bug in SKaMPI and
node times equal to 0 are at least suspicious.
- Message lengths and bandwidths now are shown in bytes or Kbytes
depending on the amounts: Less than 1024 bytes respectively bytes/s
are shown in bytes respectively bytes/s, more than or equal to 1024
bytes resp. bytes/s in Kbytes resp. Kbytes/s. Within comparison
tables the output depends on the minimum numbers for each column of
message length or bandwidth.
- Other minor changes of output have been done both at the console
and in the report file to improve readability and the like.