Re: embedding a manifest into an executable file on the fly

Tech-Archive recommends: Repair Windows Errors & Optimize Windows Performance

mayayana escribió:

That sounds like a couple of different issues:
permissions and GUI. But GUI -- getting the
visual style -- is what you're concerned about?
So it's a given that you must have an external
manifest in XP and internal in Vista/7?

I don't want to touch permissions at all, I just want to enable the visual styles (only under Vista and not in XP).

I think it would have to be a very complex job. The
way that resources are set up in a PE file is that
the Section Table (near the beginning of the file)
points to the beginning offset --
and identifies the size of -- the Resource Table. The
Resource Table then has a complex tree structure,
with pointers that are used to find specific resource
types -- and specific items of those types -- spread
out in ever-expanding roots of the tree. (For
instance, icons are one type, VS_VERSION_INFO
is another, strings a third, etc. I think a manifest
is actually a newly defined type. So what you're talking
about is not just packing the manifest into the file.
You'd also need to edit the Section Table and somehow
recalculate pointers along the levels of the tree. To reach
just a single icon requires walking several pointers down
into the Resource Table. Each pointer is a numeric offset.
A pointer either points to the next pointer or, with the
final pointer, points to the actual resource bytes. It's
an interconnected web.

OK, but I'll do it if I have the clear information of what I need to do.

Why not just ship both EXEs and then install according
to OS version?

Because I want to make it ready for general purpose. And also to avoid this additional task of having to prepare more than one exe for the package for every program.

Sometimes I prefer to work a lot in complex things, with the purpose of simplifying and automating the tasks later.

I still don't know how the AV's will consider it when my installer not only decompress an exe, but adds some resources on the fly. I hope that it's not a problem but I'm not sure.

> (And don't forget to check for the *actual*
OS version rather than just believing what GetVersionEx
tells you. :)

I know nothing about "the problem" of GetVersionEx...
.

Relevant Pages

  • Re: [PATCH] x86, ioremap: use %pR in printk
    ... the first patch introduces a "small" resource pointer printout ... The output format is - minimum width is 8. ... return string(buf, end, sym, field_width, precision, flags); ...
    (Linux-Kernel)
  • Re: Summary of resource management discussion
    ... Resource management generally works by apply resource controls over a -group- ... The patches introduce a new pointer in task_struct, ... and a new structure 'struct container'. ... one/more/all resource controllers to a hierarchy). ...
    (Linux-Kernel)
  • Re: portable finalizer
    ... is a pointer to a C structure. ... The user has to be able to close the resource explicitly (eg in ... so there is a destroy method: ... (setf pointer nil)))) ...
    (comp.lang.lisp)
  • Re: portable finalizer
    ... is a pointer to a C structure. ... The user has to be able to close the resource explicitly (eg in ... so there is a destroy method: ... (setf pointer nil)))) ...
    (comp.lang.lisp)
  • Re: RTFM
    ... > The GUI can also be the best interface for certain types of tasks. ... > and I found I didn't miss having a command line most of the time. ... A GUI is a resource. ...
    (comp.os.linux.misc)