Two more consts for PF

Hg-commit: 8b9da9d005b9944744d8a4c33d0df55482baacee

Mark the source array as const in PFApply

Hg-commit: 50c3d74c93b21118d7a95851e1e39567c1c54848

merge empty IS fix from 3.1

Hg-commit: ed1ad5aa2700e881dbc7b790b3c0a5012fd44073

Missed the empty index set case

Hg-commit: 10240efb24ab68ad285699affb77ec780284a7a9

merge ISComplement fix from 3.1

Hg-commit: 19be5646529735b64ba4af6b210d7844317d9f59

Fix ISComplement when the input IS has duplicates.

The old version allocated nmax-n space, which is not sufficient for
nmin=0 with duplicate entries (memory corruption). It is also way too
much spa

Fix ISComplement when the input IS has duplicates.

The old version allocated nmax-n space, which is not sufficient for
nmin=0 with duplicate entries (memory corruption). It is also way too
much space whan nmin is large (potential space bottleneck). The new
version scans the array to find the number of unique values, then
allocates exactly the correct amount of space (nmax-nmin-unique) for the
complement, which is passed off with ISCreateGeneralNC.

Hg-commit: 5df35d84b28ba16e9d6913ebc9f54c9604e3f2d0

show more ...

better version of PetscAMSPublishAll

Hg-commit: 782c2a976ec07d13b492bc3c4a0989c9b56cd366

Doc fix

Hg-commit: 3a931bbd8990a6f7a8b6063622130a26e3a1515f

basic amspublishall. Release it is wrong and need to publish after the type is set.

Hg-commit: 00a22ad47a7a0ffcd91c66290fefb82fce41e498

basic object publishing

Hg-commit: 0bce7cb999039c37cb50c4345a6b2f0467cf0a9d

handle attaching to AMS when no communicator exists cleanly

Hg-commit: 6565f5dcd6bd1b163afb288e69b0887d13e3a48f

simple AMSPublishAll finally works

Hg-commit: aeee3e368e2a942606f76abdcc0348e4d3b9f981

now that AMS publisher can access get a bit more of the built in AMS server working

Hg-commit: 22d1028ba0e1da07c542bf1a83b8ae6b297c7790

Formatting in VecScatterView

Hg-commit: bb7aecfb6e8d3d693f9a713b39434ea1759c943e

fixed pthread to support single precision 64 bit int windows etc

Hg-commit: 72a65071199249112327e74df3d050626ea6fb7c

Free3 frees in reverse order so don't need temporaries here

Hg-commit: e61a6b9260b4ad798350ed194132d244ae42b9fe

fixed unused variables in vpscat.h
started PETSc web server code

Hg-commit: 4a8175e1d4491aa088d0d5fc2e3bb047994f410a

updates to CUDA code for parallel vec scatters

Hg-commit: 8ea0e82fa72e5467ff14f024b348c1d95a1e0e11

start fixes for cuda scatter
fix uninitialized variable in send in ascii flush

Hg-commit: 095d62e648ed0795f6e37bee51ccd6d1a68a51e0

Fix case of Malloc2+Malloc1 being freed by Free3.

Also extract pointers from the struct so that the macro doesn't access
memory after it has been freed (when in debug mode with separate mallocs).

H

Fix case of Malloc2+Malloc1 being freed by Free3.

Also extract pointers from the struct so that the macro doesn't access
memory after it has been freed (when in debug mode with separate mallocs).

Hg-commit: bc7428ff6a678f0358d06e26002158f5122c0c67

show more ...

Some VTK improvements

Hg-commit: 73105108c2c524c138f3393b3f644a5444844a0d

Lots of inconsistent PETSC*_DLLEXPORT

Hg-commit: 17c71edbc1aa078557adacd3101bfb9934ffc5e2

_p_PetscLayout -> _n_PetscLayout, since it is not a PetscObject

Hg-commit: 379866b5e2ddbf259b3411c087ee403b6cd4a01d

Consistency checking for __FUNCT__ when __func__ or __FUNCTION__ are available.

Uses the compiler definitions in CHKERRQ/SETERRQ when available to get
better traces through user code that doesn't ma

Consistency checking for __FUNCT__ when __func__ or __FUNCTION__ are available.

Uses the compiler definitions in CHKERRQ/SETERRQ when available to get
better traces through user code that doesn't manually defined __FUNCT__.

Fix lots of places that __FUNCT__ was used inconsistently.

Hg-commit: fdbdc93647ff934a12906fd51bc39302de07ae77

show more ...

better message in check of positive definite shift for ILU

Hg-commit: 571800d0bf679ff93432b36bb77cdcb2ae4c4eea