You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I think there are a few errors in nc_def_var_extra with parallel builds.
I am trying to enable the parallel compression feature, and it is always returning an NC_EINVAL from line 673. This happens whether I have shuffle enabled or not.
#ifndef HDF5_SUPPORTS_PAR_FILTERS
if (h5->parallel == NC_TRUE)
if (nclistlength(var->filters) > 0 || fletcher32 || shuffle)
return NC_EINVAL;
#endif
I think the code is checking shuffle and it should be checking *shuffle (Or, check that shuffle is non-NULL and then check the value and return if the value is non-zero)
Note that the call to this routine from NC4_def_var_deflate is:
And shuffle is an integer parameter to NC4_def_var_deflate, so &shuffle will always be a non-NULL address, so the test in nc_def_var_extra
The overall behavior is probably the same since this is inside a block only active if the underlying HDF5 doesn't support parallel compression, so eventually the code will return with NC_EINVAL, but it shouldn't be returning that at this point.
The text was updated successfully, but these errors were encountered:
This may be related to #1713.
I think there are a few errors in
nc_def_var_extra
with parallel builds.I am trying to enable the parallel compression feature, and it is always returning an
NC_EINVAL
from line 673. This happens whether I haveshuffle
enabled or not.I think the code is checking
shuffle
and it should be checking*shuffle
(Or, check that shuffle is non-NULL and then check the value and return if the value is non-zero)Note that the call to this routine from
NC4_def_var_deflate
is:And
shuffle
is an integer parameter toNC4_def_var_deflate
, so&shuffle
will always be a non-NULL address, so the test innc_def_var_extra
The overall behavior is probably the same since this is inside a block only active if the underlying HDF5 doesn't support parallel compression, so eventually the code will return with
NC_EINVAL
, but it shouldn't be returning that at this point.The text was updated successfully, but these errors were encountered: