Unexpected behaviour in for loop

[thomasfrederikse]

I got some unexpected outcomes while running GSW-julia inside a for loop. If I run the same GSW function with the same input arguments consecutively, the results change between iterations.

A minimal working example that produces the unexpected behaviour:

for i in 1:10
   println(gsw_p_from_z(-780.0,32.0))
end

I expected that on every line the right answer 786.9634207624217 would be printed, but I get as output:

786.9634207624217
786.9634207624217
1821.1936700582019
1821.1990180813314
1821.1990181089222
1821.1990181089227
1821.1990181089227
1821.1990181089227
1821.1990181089227
1821.1990181089227

I'm using GibbsSeaWater.jl version 0.10.0 on Julia 1.4.1. This behavior is not limited to 'gsw_p_from_z', but occurs in other functions as well.

[dankelley]

In case it's of any help, the first number is the same as what I get in R, viz.

> library(gsw)
> options(digits=16)
> for (i in 1:10)
+    print(gsw_p_from_z(-780.0,32.0))
[1] 786.9634207624217
[1] 786.9634207624217
[1] 786.9634207624217
[1] 786.9634207624217
[1] 786.9634207624217
[1] 786.9634207624217
[1] 786.9634207624217
[1] 786.9634207624217
[1] 786.9634207624217
[1] 786.9634207624217

[thomasfrederikse]

Yes, the output you show is what I'd expect in Julia as well.

[kouketsu]

Thanks @thomasfrederikse and @dankelley.
I think the wrapper script "gen_gswteos10.jl" was old and did not work well with GSW-C#95a9cae.
So I updated it (and add wrapper functions for "gsw_p_from_z").

Please try master branch (through pkg>update GibbsSeaWater#master).

[thomasfrederikse]

Thanks @kouketsu! Just updated the code, and now I get the correct results. Again, thanks for the quick fix!

[Alexander-Barth]

@kouketsu should we make a new release?

[kouketsu]

@Alexander-Barth yes.