On 25/04/21 07:20 PM, Drew Parsons wrote: > Wait no, it doesn't work that way. Removing packages has to be requested if > it built in the past but no longer does so. I've had to do that > progressively with pygalmesh as it grows to exceed memory limitations of > non-intel systems. > > But if it never built in the first place then there is no problem. There's > no package to remove. The build failures don't hold up the package if > they've been failing from the beginning. > > For that reason, better to have failing build logs, I think. Once xsimd > fixes its handling of other arches, it won't have to renege on that. Ah, yes that makes sense - I got that wrong. > I think your arguments are stronger than mine on this point. Better to > change xtensor. > It was only reintroduced last month, so now is a good time to change it. > I'll get that in train. ACK. @Julien, I did another review and really liked the changes you made to the package. I did some minor changes in copyright and uploaded. Thanks a lot for having worked on this :-)) Nilesh
Attachment:
signature.asc
Description: PGP signature