[seqfan] insufficient gp/pari precision gives incorrect results

[Georgi Guninski]

insufficient gp/pari precision gives incorrect results.
probably this feature should be documented.
i got hit by it in mass pari computations.

example session:
A078976 Numerator of n-th convergent to e^(2/3)

parisize = 8000000, primelimit = 500000
? a(n)=component(component(contfracpnqn(contfrac(exp(2/3), n)), 1), 1)
? n=150
%1 = 150
? a1=a(n)
%2 = 8775965436819116582
? default(realprecision,300)
%3 = 300
? a1=a(n)
%4 =
1662089248224191266325975305441010031166997357354140605757488607965909355236108828684398094504604203059026215637276539867276263951376657621927749105770538
? default(realprecision,3000)
%5 = 3000
? a1=a(n)
%6 =
666280308971067538803113985595492989367403873213980495571359635598141444179643386508726043481092812835676179552224737408022484748450993933584137176738055064772678238280831940888731615490091677245778769175071
? default(realprecision,30000)
%7 = 30000
? a1=a(n)
%8 =
666280308971067538803113985595492989367403873213980495571359635598141444179643386508726043481092812835676179552224737408022484748450993933584137176738055064772678238280831940888731615490091677245778769175071