Controlling the expansion of variable expressions using a global
variable and a VARE flag was inconsistent.
Converting the global variable into a flag had to prerequisites:
1. The unintended duplicate variable assignment had to be fixed, as
done in parse.c 1.520 from 2020-12-27. Without this fix, it would have
been necessary to add more flags to Var_Exists and Var_SetWithFlags, and
this would have become too complex.
2. There had to be a unit test demonstrating that VARE_KEEP_DOLLAR only
applies to the top-level expression and is not passed to the
subexpressions, while VARE_KEEP_UNDEF applies to all subexpressions as
well. This test is in var-op-expand.mk 1.10 from 2020-12-28, at least
for the ':@word@' modifier. In ParseModifierPartSubst, VARE_KEEP_UNDEF
is not passed down either, in the same way.
At that point, the expression can never be varUndefined. At the
beginning of ParseVarnameLong, the expression is initialized to a simple
empty string, and that string is only ever converted to varUndefined at
the very end of Var_Parse.
Back in 1993, the variables in a context were stored in a linked list.
Searching such a list indeed required literally thousands of calls to
strcmp. In make.h 1.22 from 1999-09-15, the linked list was replaced
with a hash table, requiring much fewer string comparisons. Since then,
the rationale doesn't apply anymore.
This allows the -q option to distinguish errors from out-of-date
targets. Granted, it's an edge case but it should be solved
consistently anyway.
The majority of cases in which make exits with exit status 1, even in -q
mode, is when there are parse errors. These have been kept as-is for
now as they affect many of the unit tests.
The technical errors, on the other hand, occur so rarely that it's hard
to write reliable tests for them that fail consistently on all platforms
supported by make.
Extracting the character-level details makes the essence of Var_Subst
visible in the code, which is to iterate over the given text, handling a
few types of tokens.
The many constants were invented because at that time I didn't quite
understand the actual outcomes of Var_Parse that need to be
distinguished. There are only a few:
(1) Errors, whether they are parse errors, or evaluation errors or
undefined variables. The old constants VPR_PARSE_MSG and
VPR_UNDEF_MSG are merged into VPR_ERR.
(2) Undefined expressions in a situation in which they are allowed.
Previously the documentation for VPR_UNDEF_SILENT talked about
undefined expressions in situations where they were not allowed.
That case is fully covered by VPR_ERR instead.
(3) Errors that are silently ignored. These are probably bugs.
(4) Everything went fine, the expression has a defined value.
Right now, Var_Subst always returns VPR_OK, even if there had been parse
errors or evaluation errors. If that is no longer true, the errors will
be reported properly.
Since make uses vfork if available, re-exporting the variables happens
in the address space of the main process anyway, so there is no point in
mentioning anything about "our client process" anywhere.
Previously, running lint mode didn't define MAKE_RCSID at all, which
resulted in a syntax error.
While here, reduced the indentation and nesting of the preprocessor
directives.
When running lint(1) on the code, it defines the preprocessor macro
"lint" to 1, which generated a syntax error in the declaration "Boolean
lint", as that became "Boolean 1".
This macro was supposed to return a boolean expression all the time, it
just hadn't been implemented this way. This resulted in wrong output
for the test sh-flags, in compilation modes -DUSE_UCHAR_BOOLEAN and
-DUSE_CHAR_BOOLEAN, since in ParseCommandFlags, the expression
DEBUG(LOUD) didn't fit into a boolean.
Since make doesn't support variable names containing spaces, this edge
case is not enough reason to stop this feature. Having multiple
variable names as arguments nicely aligns with other directives such as
.for and .export.
Previously, mmapped files didn't always have the final newline added.
Only those that ended at a page boundary did.
This confused ParseRawLine, which assumed (and since parse.c 1.510 from
moments ago also asserted) that every line ends with a newline, which
allows the code to assume that after a backslash, there is at least one
other character in the buffer, thereby preventing an out-of-bounds read.
This bug had been there at least since parse.c 1.170 from 2010-12-25
04:57:07, maybe even earlier, I didn't check.
Now line_end always points to the trailing newline, which allows
ParseGetLine to overwrite that character to end the string.
