pre or post condition
jacques silberstein
Cordialement Jacques Silberstein
In Vision2.
when calling:
set_column_alignment (an_alignment: EV_TEXT_ALIGNMENT; a_column:
INTEGER_32)
-- Align the text of column `a_column` to
`an_alignment`
-- The first column must stay as left aligned as MSDN
-- states that the first column can only be set as
left aligned
-- for Win32.
-- (from EV_MULTI_COLUMN_LIST)
require -- from TOOLS_COLLECTION_PRIMITIVE
True
require -- from EV_MULTI_COLUMN_LIST
not_destroyed: not is_destroyed
a_column_within_range: a_column > 1 and a_column
<= column_count
alignment_not_void: an_alignment /= Void
do
implementation.set_column_alignment (an_alignment,
a_column)
ensure -- from EV_MULTI_COLUMN_LIST
column_alignment_assigned: column_alignment
(a_column).is_equal (an_alignment)
end
I get the message:
call Stack
a_column_within_range: PRECONDITION_VIOLATION raised
Without assertion check, over GTK the right alignment work properly.
Conceptualy, it is mot unsfull to build a complexe multi-layer in order to be independant if the platform dependent layer.
Then to solutions:
1) a_column_within_range: a_column >= 1 and a_column <= column_count
and, at msw implementation the request on a right aligment has no effect.
2)low_column_idx: (platform = windows implies a_column > 1
low_column_idx: (platform /= windows implies a_column >= 1
The second solution seem a litle bid ugly but...
Liberty Lover
Your analysis is correct, and your screenshot confirms it perfectly. The watch panel shows that a_column >= 1 and a_column <= column_count evaluates to True - that's the precondition that should be there.
Your Solution 1 is the right approach from a Design by Contract perspective:
a_column_within_range: a_column >= 1 and a_column <= column_count
This preserves EiffelVision2's purpose: platform-independent GUI programming. Code that works on GTK should not crash on Windows because of a platform quirk baked into a precondition.
Solution 2 (platform-specific preconditions) breaks the abstraction - it forces every caller to know about Win32 ListView internals, which defeats the point of the abstraction layer.
Recommendation: Report this to ISE as a library issue. The fix is straightforward:
1. Relax the precondition to a_column >= 1
2. Have the Win32 implementation silently ignore column 1 alignment requests
Workaround until fixed:
if a_column > 1 then
my_list.set_column_alignment (an_alignment, a_column)
end
Best,
Larry
--
You received this message because you are subscribed to the Google Groups "Eiffel Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to eiffel-users...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/eiffel-users/c6b2ae3f-2871-41b3-ad64-62606acbaa8c%40abstraction.ch.
Ulrich Windl
Just wondering: Should the Windows mis-feature be mapped in the postcondition instead? Like "ensure column_1_is_left_aligned:"
Ulrich
12.01.2026 17:55:26 Liberty Lover <rix....@gmail.com>:
> --
> You received this message because you are subscribed to the Google Groups "Eiffel Users" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to eiffel-users...@googlegroups.com.