It was a combination for me, depending on the lecturer. I had some papers where up to 15% of the paper purely depended on output being 100% correct (you lost percentages each time it failed, and got 0% if it didn't work in the end).
I also had some papers where your grade was based around functionality, where there were several milestones, so as long as you got it partially right, you got some marks. For instance, in one paper, we had to implement RIPv2, but we got partial marks if we only got parts of the spec working.
Then there were papers where you had a final goal, but even if your code didn't even compile, they'd take a look at it and if you were on track, you'd get partial marks. For example, we had a paper where we had to control a helicopter reading data from a rotary encoder and other inputs, and use PID control. We had our PID control backwards so the thing wouldn't even fly, but the code was otherwise correct.
I also had some papers where your grade was based around functionality, where there were several milestones, so as long as you got it partially right, you got some marks. For instance, in one paper, we had to implement RIPv2, but we got partial marks if we only got parts of the spec working.
Then there were papers where you had a final goal, but even if your code didn't even compile, they'd take a look at it and if you were on track, you'd get partial marks. For example, we had a paper where we had to control a helicopter reading data from a rotary encoder and other inputs, and use PID control. We had our PID control backwards so the thing wouldn't even fly, but the code was otherwise correct.