Last PE section does not have to be padded up to IMAGE_NT_HEADERS.FileHeader.FileAlignment. Would FASM rather be able to do this size optimization and not pad up last section (and of course adjust IMAGE_SECTION_HEADER.SizeOfRawData to reflect the change as well)?

Personally (in general), I think somehow such kind of options should be controllable via options, including the new test instruction optimization.

privalov still refuses to allow command line options.

Why? They could be optional, so no backward compatibility breaks, and newbies still could use "fasm <source> <destination>".

When did I say that I refuse? I plan to implement some switches than don't affect the output code, like the switch for setting the upper limit for memory usage. But all setting that affect the output are in fasm done using the directives in the source.

But you are not right about the section alignment, this is what PE/COFF docs from MS say about the SizeOfRawData field:

Try it out

Here is program from FASM\EXAMPLES\HELLO

This technique can be achieved using PE realigner program, so not really necessary for FASM to have it. Just bringing to your attention that it is indeed true. Very minor.

We cannot ignore the official specification just because it works.

I'm with privalov. Things have worked that broke on later windows versions (certain PE aspects, no-import PE's, port I/O, ring0/admin hacks, ...)

privalov: so where is option to generate only result of preprocessing? It would be very useful and as I know, it is very easy to do (just save some buffer to file and exit)

Why are you asking about it in this thread?