Re: Lilypond error behaviour

[Kieren MacMillan]

Hi all,

On Apr 17, 2016, at 8:15 AM, David Kastrup <address@hidden> wrote:
> why can one now consider LilyPond as having succeeded in compiling the file?

If no final output file (e.g., PDF) exists before compilation begins, a final 
output file does exist after compilation ends, most users would understand the 
aggregate error to be non-“fatal”. To me — and, I would offer, to most users 
I’ve ever worked with, programmed for, or watched in action — “fatal”, in this 
kind of context, means the application in question had to terminate before 
*any* [not *all*] meaningful/useful/concrete/verifiable output could be 
achieved. Surely, if a database made a successful SQL transaction, but the UI 
couldn’t return to the main menu because of some subsequent navigation error, 
it would be misleading (not to mention unnecessarily alarming) to say that a 
“fatal” error had occurred.

That being said, if (as in David K’s example) one or more sub-compilations 
fails “fatally” — in other words, it would have resulted in no 
meaningful/useful/concrete/verifiable output if it were the sole compilation 
task — then the log should reflect that in some way: it should indicate that 
*some* meaningful/useful/concrete/verifiable output had been achieved, but 
possibly not *all* of it as expected by the user.

Nobody ever said good error handling and reporting was easy.  =)

Cheers,
Kieren.
________________________________

Kieren MacMillan, composer
‣ website: www.kierenmacmillan.info
‣ email: address@hidden