create new tag
view all tags

Comparing the pdfi branch with Ghostscript "master" on the cluster

Importantly: this relies on the ability to run the pdfi interpreter "inside" Ghostscript (i.e. calling it from Postscript). Comparing gpdf to gs will require rather non-trivial cluster surgery.

Also important: this is a first pass at this. The process probably can, and probably will be streamlined in the future.

Firstly, the code to test in pdfi must exist in a git commit and, following these steps, will be compared to the tip of the ghostpdl "master" branch. In general, therefore, it would be best to ensure than your pdfi branch is up to date with master before doing the tests (unless you have a specific need for it*not* to be up to date, of course).

On the pdfi branch, create a temporary "sacrificial" commit (after the commit you want tested), changing the Postscript in Resource/Init/pdf_main.ps to forcibly use the pdfi interrpeter instead of the Postscript implementation. In "/runpdf":

- /NEWPDF where {/NEWPDF get not} {true} ifelse

+ //false

With that commit tag that commit - I'd like to use a naming pattern for these tags "gpdf_gs_text_filter_xyz" (where "xyz" is an incrementing 3 digit number - gpdf_gs_text_filter_000 already exists!). That way it will be trivial to clear them out when we want to.

Once all that is done, do:

git push --tags golden gpdf_gs_text_filter_xyz

where "golden" is whatever you call the main ghostpdl repo on casper, and "gpdf_gs_text_filter_xyz" is the tag you created above - this will push the tag but not update the branch.

Now, log into casper, and become the "regression" user. Edit the file:


So that the "rev" fields reference the tags you created above.

Then run, in the /home/regression/cluster directory, run:

./enqueueAuto.pl auto/chrisl/gpdf_gs_text_filter

The cluster results will land in:


and the bmpcmp in:


Note, the previous results will be overwritten when a new run is completed. If you want to keep them safe, ask Robin or I to copy them somewhere. They do take up considerable space, so I'd prefer to keep them longer term only when it's necessary.

-- Chris Liddell - 2021-03-12


Edit | Attach | Watch | Print version | History: r3 < r2 < r1 | Backlinks | Raw View | Raw edit | More topic actions
Topic revision: r3 - 2021-03-12 - ChrisLiddell
This site is powered by the TWiki collaboration platform Powered by PerlCopyright 2014 Artifex Software Inc