Reset page vertical space on top?

Question

This is one of the cases, where I've tried for some time now to come up with a minimal example that will demonstrate the problem, but I can't. All I can provide is the problematic behavior of the master document, so I hope you'll bear with me.

Essentially, I use titlesec for custom headings; and pdfpages to include PDF pages. I have a structure like:

\includepdf{...}
\cleardoublepage

\tracingpages=1
mysection*{Section X} \tracingpages=0

% rest of content...

When it's like this, all is good, and I can confirm that all sections are positioned at the same location on the page (EDIT: a good explanation of the \tracingpages log is in #60213 - How can I create a chapter with graphics only divided into different sections?):

$ grep '^%' mydocument.log

%% goal height=591.5302, max depth=5.5
% t=14.4 g=591.5302 b=10000 p=0 c=100000#
% t=-10.76009 g=591.5302 b=10000 p=-300 c=100000#
% t=35.31354 plus 0.15234 g=591.5302 b=10000 p=0 c=100000#
%% goal height=591.5302, max depth=5.5
% t=0.0 g=591.5302 b=10000 p=0 c=100000#
% t=14.4 g=591.5302 b=10000 p=0 c=100000#
% t=-10.76009 g=591.5302 b=10000 p=-300 c=100000#
% t=35.31354 plus 0.15234 g=591.5302 b=10000 p=0 c=100000#
%% goal height=591.5302, max depth=5.5
% t=14.4 g=591.5302 b=10000 p=0 c=100000#
% t=-10.76009 g=591.5302 b=10000 p=-300 c=100000#
% t=35.31354 plus 0.15234 g=591.5302 b=10000 p=0 c=100000#
%% goal height=591.5302, max depth=5.5
% t=14.4 g=591.5302 b=10000 p=0 c=100000#
% t=-10.76009 g=591.5302 b=10000 p=-300 c=100000#
% t=35.31354 plus 0.15234 g=591.5302 b=10000 p=0 c=100000#

So, all sections end at top position t=35.31354, and they are indeed aligned fine.

The problem is, sometimes \cleardoublepage inserts a blank page twice. If the included pdf ends on odd page, then \cleardoublepage correctly inserts a blank page, and new section starts on next odd. If the included pdf ends on even page, then I get a blank on (next) odd page; \cleardoublepage apparently picks this up, and inserts one more as even - and the new section starts on (next+2) odd page.

So if I remove \cleardoublepage from a spot like that, then the next "Section X" starts at the next page after PDF is finished, and is on an odd page, as expected - but now the title is slightly misplaced vertically ( :/ ) ; here is what the log reports:

%% goal height=591.5302, max depth=5.5
% t=14.4 g=591.5302 b=10000 p=0 c=100000#
% t=-10.76009 g=591.5302 b=10000 p=-300 c=100000#
% t=35.31354 plus 0.15234 g=591.5302 b=10000 p=0 c=100000#
%% goal height=591.5302, max depth=5.5
% t=0.0 g=591.5302 b=10000 p=0 c=100000#
% t=14.4 g=591.5302 b=10000 p=0 c=100000#
% t=-10.76009 g=591.5302 b=10000 p=-300 c=100000#
% t=35.31354 plus 0.15234 g=591.5302 b=10000 p=0 c=100000#
# NOTE NO 'goal height' HERE!!
% t=40.68004 plus 4.0 minus 6.0 g=591.5302 b=10000 p=0 c=100000#
% t=15.51996 plus 4.0 minus 6.0 g=591.5302 b=10000 p=-300 c=100000#
% t=61.59358 plus 4.15234 minus 6.0 g=591.5302 b=10000 p=0 c=100000#
%% goal height=591.5302, max depth=5.5
% t=14.4 g=591.5302 b=10000 p=0 c=100000#
% t=-10.76009 g=591.5302 b=10000 p=-300 c=100000#
% t=35.31354 plus 0.15234 g=591.5302 b=10000 p=0 c=100000#

So obviously the third listed part is the problematic one:

• There is no %% goal height - as if a new page never started;
• The first top position reported is 40.68004 plus 4.0 minus 6.0, unlike the others being 14.4
• The top then seems to "reset" itself to 15.51996 - but it needs to be on -10.76009 to align like the others.

To me, the problematic part looks like a state, as if we're "currently" on top of a blank page; yet latex still thinks of it as not a new page.

Note that I'm not keen on using \vspace{-Xpt} or similar to manually correct, since I'd have to tune the value by hand, and the master document takes several minutes to build... (and what guarantee do I have that all eventual problematic sections will be misplaced by the same ammount?)

---

So is there a command I could use, that could "reset" the "current page", so that the "top position" would be reset to a default value (_{as a side effect, I'd expect if you use such a command on a page that is typeset halfways, that the "cursor" would essentially go up, and typeset over existing contents})

---

PS: As an experiment, I tried to get the last statements from \newpage:

$ texdef -t latex -c book \newpage -f
\newpage is defined by (La)TeX.

\newpage:
macro:->\if@noskipsec \ifx \@nodocument \relax \leavevmode \global \@noskipsecfalse \fi \fi \if@inlabel \leavevmode \global \@inlabelfalse \fi \if@nobreak \@nobreakfalse \everypar {}\fi \par \vfil \penalty -\@M 

... and add them instead of the problematic \cleardoublepage, in hope that they might "reset" the page without breaking it - however, even just \penalty -\@M on the problematic spot inserts a new blank page (and also \leavevmode at the same position does nothing to remedy the position)

Answer 1

Ok, FWIW (without a proper MWE), I fixed my problem with \unskip (as I heard somewhere it resets skips), and \vspace* - but the question for a command that "resets the top" still stands.

First, I tried inserting an \unskip instead of the problematic \cleardoublepage; that produced "problematic" page section measurements of:

% t=28.68004 g=591.5302 b=10000 p=0 c=100000#
% t=3.51996 g=591.5302 b=10000 p=-300 c=100000#
% t=49.59358 plus 0.15234 g=591.5302 b=10000 p=0 c=100000#

So, I get a 3.51996 (nicely, without plus or minus components); and for proper alignment, I should have -10.76009; so I take I need 3.51996-(-10.76009) offset upwards, so I simply enter this sequence instead of the problematic \cleardoublepage:

...
\unskip
\vspace*{-14.28pt}
...

... and now the (final) page measurements are:

% t=14.40004 g=591.5302 b=10000 p=0 c=100000#
% t=-10.76004 g=591.5302 b=10000 p=-300 c=100000#
% t=35.31358 plus 0.15234 g=591.5302 b=10000 p=0 c=100000#

There is but 0.0004 pt difference, and visually I cannot see it - so I'm ok with this.

Another thing is that the mentioned \mysection is actually a titlesec custom sectioning command; and another thing I didn't mention above, is that the sequence more accurately looks like this:

\newcommand{\SetMyVar}[1]{%
\def\MyVar{#1} % 
}

...

\includepdf{...}
\cleardoublepage

\SetMyVar{1}
\tracingpages=1
mysection*{Section X} \tracingpages=0

....

(yes, the \SetMyVar macro is that trivial - it's simply for syntactic consistency with other macros I use)

Now this is the trick - if I replaced the \cleardoublepage "as is" - like this:

\includepdf{...}
\unskip
\vspace*{-14.28pt}

\SetMyVar{1}
\tracingpages=1
mysection*{Section X} \tracingpages=0

... I could in fact not see any expected displacement in the logfile! However, when I then placed the \cleardoublepage replacement after the \SetMyVar macro call:

\includepdf{...}

\SetMyVar{1}
\unskip
\vspace*{-14.28pt}
\tracingpages=1
mysection*{Section X} \tracingpages=0

... then finally it started working - and I got the (final) page measurements mentioned above.

I still wish I could reset with a command, instead of "compensating" "manually" - but as long as this approach keeps on working, I'll be happy :)

Hope this may help someone,
Cheers!