Errata

In hg 1.5 (and presumably the version used for the example on p192) qnew incorporates any changes in the workspace into the created patch, and the -f option is deprecated.

The fourth bullet point is incorrect, if there is a positive guard that isn't specified, the patch isn't applied. The statement on the previous page that negative guards take precedence over positive is misleading IMHO.
An alternative description of the rules might be:
# If the patch has any positive guards that are not selected, then the patch will be skipped,
# If the patch has any negative guards that are selected, then patch will be skipped,
# Otherwise the patch will be applied.

Test script (its a windows batch, deal):
mkdir qguardrepo
cd qguardrepo
hg init
echo [extensions] >> .hg/hgrc
echo hgext.mq = >> .hg/hgrc

hg qinit

for %%i in (A, B, C, D, E, F, G, H) do (
echo %%i > %%i.txt
hg add %%i.txt
hg qnew %%i.patch
)

hg qguard -- B.patch +one +three
hg qguard -- C.patch +three
hg qguard -- D.patch +one -three
hg qguard -- E.patch -three
hg qguard -- F.patch +one -two
hg qguard -- G.patch -two
hg qguard -- H.patch -three +four

hg qselect one two
hg qpop -a
hg qpush -a
dir /b
hg --version

The output from the last three commands is:
$ hg qpush -a
applying A.patch
applying B.patch
skipping C.patch - guarded by ['+three']
applying D.patch
applying E.patch
skipping F.patch - guarded by '-two'
skipping G.patch - guarded by '-two'
skipping H.patch - guarded by ['+four']
now at: E.patch

$ dir /b
.hg
A.txt
B.txt
D.txt
E.txt

$ hg --version
Mercurial Distributed SCM (version 1.5.1)

PS. It would be nice to have a preview for the errata submissions...

The output from both command sequences is unexpected/incorrect.

The "cat myfile" will output all three lines, as Kevin notes for p130, the working directory is still at r2.

The "hg merge" is reported as aborting due to uncommitted changes, presumably the same (not shown) change that caused the incorrect cat output above.

In fact, the merge will succeed.

As a result, the final "cat myfile" will display:
first change
third change
as one would hope the backout result to be.

If a patch has a negative guard, but it doesn't match the currently selected guard, the patch is applied.

Therefore the fourth bullet isn't correct to say the patch is skipped. (I think - the whole section left me a bit confused).

The example is not the correct example. It should show what happens when a qnew is performed with an outstanding change. Instead it is showing that creating the same patch file twice generates and error.

After performing the 'hg backout', the parent of the working directory is still 2:third change. The entire discussion at the bottom of the page "Let's think about what we expect to see..." is incorrect. The contents of myfile will contain all three lines until we 'hg merge'.

The node id in the "First Parent" box should be 7b064d8bac5e not e7639888bb2f.