I released this new version in a development line at first, because it has some extensive changes to the assembler's internal engine - they should not affect most the of sources, however it is possible that I inadvertently introduced some bugs. Also, as this revision fixes a number of bugs that could be sometimes (mis)used by the programmers, it is possible that some source will cease to assemble because of it.

The changes may not sound like much, and most of you will perhaps not even notice that something is different, but it required a lot of work to tweak the internals in order to get this working, and I'm quite satisfied that I finally managed to correct one of the serious fasm's flaws.

What has been done? I have finally implemented the 65-th precision bit for expression calculator engine, and thanks to this addition I was able to correct many of the behavior that I disliked. The new engine is able to distinguish large numbers in unsigned 64-bit range from the negative numbers in signed 64-bit range, and thus it is able to correctly detect overflows on addition and subtraction (and also "shl"). It also able to multiply and divide in the full composite range.

Another consequence is that expressions for QWORD size now behave consistently with values of other sizes. For example: With earlier fasm versions "qword" override had no such effect (since fasm was not able to distinguish that large positive number from -1 constant).

The changes also had to be applied to "fas" file format, its documentation has been updated. I also added a concise paragraph to the fasm's manual, which describes what results should you expect from the fasm's expression calculator.

> I have finally implemented the 65-th precision bit for expression calculator engine

COOOOOOOOOOOOOOL

> I release this new version

It's available but hidden: http://flatassembler.net/download.php



What's new:

* see ^^^ above
* updated calculator in the IDE too (untested)
* increased bloat by almost 1 KiB
* made 69.39 deprecated

Scroll down.

minimal test: seems to work, calculator passes my $FFFF'FFFF'FFFF'FFFF-test (while older versions securely fail it), some code compiles (more not tested), TFM is updated too (but isn't the "infinite precision" a bit too optimistic for 65 bits ?)

The use of "infinite precision" phrase there means that fasm gives the results that are the same as the ones it would give, if it operated on truly infinite precision - and if it is not able to ensure that the result is as such, it signals the overflow error.

This is designed with possibility of future improvements in mind - for example if there is fasm 2.0 with 129-bit calculator engine, it will be able to operate give the correct results for larger values, but for the expression which even 65-bit fasm is able to calculate they will both give the same results, so this will ensure compatibility. It also makes it possible to predict the results of calculations without any knowledge about what precision assembler uses internally in given implementation - as long as the implementation is capable of computing the particular expression for you, the result should be the same as if it was given with arbitrary precision implementation.

a (no longer) silent update Mar-8 ?

What's new:

* see above
* substantial optimizations reducing bloat by over 1 KiB
* split Win32 binaries into sections (I never had had virii anyway)
* deleted both 40 and 41 versions