jfbu at free.fr
Wed Jan 11 13:16:01 CET 2017
to my dismay, xelatex and lualatex seem to follow suit regarding missing \fi's or missing \endgroup's not changing exit status from 0.
I wonder if there has been in the past some discussion about this in the respective lists, because it looks very strange for "modern" binaries to issue non-usable exit stati.
My opinion is that maintaining compatibility with Makefiles from thirty years ago is illusory, shells change too. People will already have had more problems with diverging version of ``Make/make``. I expect that if some historic archive of TeX files is to be maintained, the people in charge will have taken care of archiving the full eco-system.
It is very nice to be able to use tex as is with thirty years old files and I do not disagree at all with that, (for many years I used only TeX, and I am 100% sure those files will compile identically, but since I switched to LaTeX, I am sure they don't, for example simply from changes to the babel module for French, not to mention all other packages, I have had problems with that and needed to take extra measures like preserving style or config files, despite the underlying TeX being strictly backwards compatible and LaTeX format also -- at least to 99.9999% perhaps even 100% -- but the eco-system for LaTeX is definitely not backwards compatible), but having a constraint that scripts from thirty years ago should not be broken looks like a very conservative way. Frankly, I was shocked to see that xetex and luatex aimed at maintaining 100% compatibility there where there was option for improvement. (I have not looked at the docs to see if there was an option like Philip suggests to toggle new policy for exit stati)
>From a naïve point of view here now 15-20 years later, I am also surprised by etex's decisions. I presume this may be related to the fact that OS's were more diverse back then. With Mac OS X a flavour of unix and the spread of Linux even to one's home refrigerator (not mine!), I think in 2017 there is different situation, psychologically.
Philip Taylor wrote:
> I for one would be 100% in favour of improving the utility of the return status of the *TeX engines, enabling stati of "Success", "Informational", "Warning", "Error", and "Severe (or 'fatal') Error" and so on to be easily distinguished. If this were implemented through a new command-line parameter, no existing files ("Make" or otherwise) would be affected.
> Philip Taylor
> Karl Berry wrote:
>> I confess I am not at all inclined to fiddle with etex (or pdftex) exit
>> values. Although I don't think the current behavior is perfectly logical
>> or optimal, I also don't think the benefits of changing it would be
>> worth the agony of changing something that has been stable for
>> decades. E.g., Makefiles would start to fail where they now succeed, or
>> vice versa. I don't think we should go there. --karl
More information about the tex-live