This reverts commit 7686f93f.

Change-Id: Icdf7653e9dc5a33d78381b9ea5f3fd4652b1be5a

MSAA can and will result in mixed colors which is a problem if we use
the color for the selection.

Change-Id: Ifbd6d37902069b26e714d2e3e5576fb767f9fbbf

Based on the patch on Bugzilla by
Marcos Paulo de Souza <marcos.souza.org@gmail.com>

Change-Id: I215b83894f228b1cc8908f98858b85c8d5378ddb

Change-Id: I3b0e4ecd5def28c6b9f4639b8df16a2a0ea0008c

Change-Id: Ib792a4251ed0040d10abc6a408089d7cc8e5eca8

For long columns of repeated formula, it is far faster to convert a formula
to a string and compare to see if the next cell in a column matches, than
it is to parse the formula string itself (sad but true).

This saves 29s of 131s of load time for my (admittedly large) ODS file.

Re-instate commit - with pCode now correctly intialized.

Change-Id: Ib25840e7916c3e1c681c8100678770b7ab7222ab

Whee, seems hb has changed naming, with 0.9.11 it is spelled as
quoted. newer versions have a mapping in hb-deprecated.h ...

Change-Id: I425ded33e6b9928aadca9f6988437a021fcecbfb

Change-Id: I6621529921120170bc1572dac727e9dc3b9438d5

Change-Id: I7e78e45b66ea2b74268f4e4479ebb0259eba130c

Change-Id: Iac80d74a9ec6382a232fdc2f4b798e57dc643ad3

Change-Id: Ia5fa3f15ed3dbf37aa76959073be1796c8c8c043

Change-Id: I83949b592d75899d28d9792508685e58d103ae16

Change-Id: I700c5f6799ba59eb858d1a12807dc7e37c4a3900

It seems we need mirroring.

Change-Id: Id9ad7041eb09ec773e4174cb68d34a65a2c473fe

Change-Id: I53f6a97820da12e89d567e1bc10226fa27cafd84

Change-Id: I03580779ba69a71e51e9e4e1349fe2f5d77956d5

Change-Id: I71e7e1fbab6ec2ecc525f7d637199f6f7a69f8ea

Change-Id: I67237267be5ac4071ecdb480cacf51b4a2fb4f1f

Change-Id: I25167b28397db185cdf9fe5844feecb5a16cc13e

Change-Id: Idaa9a057bb17af2642706b144544df92139a1fce

30bb1edd originally incorrectly
changed some different 16bit numbers to 32bit, then
02abb559 reverted those 16bit->32bit
changes but accidentally changed the 32bit nAryLen to 16bit but
that was 32bit before 30bb1edd

Change-Id: I36863b62418d7a3b5ccce51aef86c63c523f9598

Change-Id: I875faea495a375b998c439a0c987b2241ca47ea8

Change-Id: I01501e99787118e7a7d8259dc65ae20a2932dea7

Change-Id: I095820c3abcb8e2b47e564096f050309b22cf901

Change-Id: I968c5cd5929e2587588f0e0547352acdc19e456a

Change-Id: I11bdec73ea12e189cca41f5f275616a138cf5793

Change-Id: Ia167c532afc8ba900629b3f056021deed45185ea

Change-Id: I475b8c9bdcea120ac7ba24862000a16d9f4127c0

Change-Id: Icf3a58257c8c8e90d7e9d38f167a9b9f80d4bf32

Setting 3 separators instead of 2 does not work.

(regression from 26315384)

Change-Id: Id0b4649404b8a49a7f89a237c46106bce4e146ff

Change-Id: Ifd6acd85a9bf33742cb62327a3b8a4320f35d0f3

Change-Id: I454e29c2cdf0857d329d84000d79973489cebb41

Otherwise we never hit cache directly, only after expensive call to
ImplFindByFont.

Change-Id: If15b368feeba94c8fff8ee7cbe049fc4a2069768

Change-Id: I35e295347a046376289f5d4fd5468860d0b8f0ae

Found when investigating fdo#79228 which was caused by the ScPatternAttr
change due to the number format being applied.

Change-Id: Ieada557f5e194401a9ce255f26d71d36e0704f35

Change-Id: I99deaf80f559e28de106b639663b262544cdf60c

Because the same dialog is used when creating a new pivot table, where
the Apply and Close semantic feels a bit weird.  Also, some features of
the dialog were designed with Ok / Cancel in mind (like the destination
location).

Change-Id: I9ac889bca9f967cb97b34ab45ea563547b58d422

Change-Id: Idd97331669ff204efb568ef1c17ba7444db4abf4

Change-Id: I24f814a402d2b8af3ddf888719b78d3158c2a38f