MNT : better error handling on determining gs version #3277
Conversation
|
Note as I pointed out in #3259 this should probably be unified with the code in checkdep_ghostscript in the long term i.e. after 1.4 |
|
I agree this is not ideal, but it is fast to implement and I don't think will break anything else. |
|
Agreed, This probably should be cleaned up post 1.4, Same goes for dvipng_hack_alpha in texmanager.py |
|
heh, seems like the travis failure is unrelated (a slowdown verification in drawing invisible lines failed in py3.3) |
|
@tacaswell , this PR needs a rebase |
|
Will this avoid the test error of #3259? The reason I ask is because it would be good to clear the test errors in the automated OSX install tests: https://travis-ci.org/matplotlib/mpl_mac_testing - that would make it easier to keep track of OSX problems in the future. |
If we fail to parse the version string return (0, 0) to let the code for dealing with out-of-date versions of gs handle the case of not installed.
|
@WeatherGod rebased @matthew-brett Yes, that is the purpose of this, if it works is a different question. |
|
@jenshnielsen Can you make sure I didn't muddle up the process communication? |
|
I looks good to me. This happens on the output of the popen communication. There is some scope for some code simplification later on post 1.4.x but not now. |
MNT : better error handling on determining gs version
If we fail to parse the version string return (0, 0) to let
the code for dealing with out-of-date versions of gs handle
the case of not installed.
Partially addreses #3259