Farid Zaripov wrote:
-----Original Message-----
From: Martin Sebor [mailto:[EMAIL PROTECTED]
Sent: Thursday, November 02, 2006 12:12 AM
To: [email protected]
Subject: Re: [PATCH] RE: testsuite process helpers
[...]
+ if ('\'' == *cmd || '\"' == *cmd) {
+ char* const cmd1 = cmd + 1;
+ // search the closing quote
+ if (char* pos = strchr (cmd1, *cmd)) {
Did we decide that it was okay to use this syntax the last
time we discussed it ot or did we say we'd avoid it? (I don't
recall and am too lazy to look it up.)
Sorry, I didn't catch what the problem with syntax (I didn't found
discussion about _rw_split_cmd() function).
Could you be more specific?
The syntax I'm talking about is the one involving the declarator,
i.e., in the above the writing the if statement like this:
if (char* pos = strchr (cmd1, *cmd))
rather than like this:
char* pos = strchr (cmd1, *cmd);
if (pos)
I thought we had discussed this but I might be thinking of the
for loop scoping discussion (starting http://tinyurl.com/y4xyph).
I'm pretty sure there are older compilers that don't support the
more concise form so I think we should continue to avoid relying
on it or decide that the syntax will be required in order to
support stdcxx. I don't think we ever came to a decision on the
for loop question either. I think it would make sense to put
together a list of modern language features currently treated
by stdcxx as optional (member templates, partial specialization,
etc.) that we will agree to assume robust support for from the
compiler. But I want that to be a conscious decision, not
something we allow to creep in without everyone's full
understanding of the implications.
[...]
Just curious: why is ERROR_INVALID_NAME not handled in _rw_map_errno?
Hmm, I get the map table from the sources of the MSVC crt (function
_dosmaperr).
I've already added handling of the ERROR_INVALID_NAME to the
_rw_map_errno.
Got it!
[...]
Because ERROR_INVALID_HANDLE by default is mapped to EBADF, but in
this
case ECHILD should be returned.
I see. Thanks.
Martin
