18.11.2020 17:49, Dimitry Sibiryakov wrote:
How long pointer returned from IStatus::getErrors() is supposed to be valid?
I think I found a weird case. In Y-valve there is code like this:
status.setErrors(status.getErrors());
During such self assignment returned pointer can be invalid
On 12/9/20 2:03 PM, Dimitry Sibiryakov wrote:
18.11.2020 17:49, Dimitry Sibiryakov wrote:
How long pointer returned from IStatus::getErrors() is supposed to
be valid?
I think I found a weird case. In Y-valve there is code like this:
status.setErrors(status.getErrors());
During such
On 12/8/20 8:20 PM, Dimitry Sibiryakov wrote:
08.12.2020 17:36, Alex Peshkoff via Firebird-devel wrote:
Should a custom implementation of IStatus recognize and ignore a
special case when GDS error code is set to zero using parameter (
isc_arg_gds, 0, isc_arg_end } or error with zero code is a
09.12.2020 12:21, Alex Peshkoff via Firebird-devel wrote:
During such self assignment returned pointer can be invalidated by setErrors() code
before using input parameter which may cause crash or wrong result.
Builtin implementation works fine with your particular sample.
Not quite so.
09.12.2020 12:23, Alex Peshkoff via Firebird-devel wrote:
That depends upon getState() internals, is not it? :)
Yes. Firebird implementation recognizes zero error code in getState() so I wonder if it
is a specs' requirement or such recognition can be performed in other places.
--
WBR, SD
On 12/9/20 2:28 PM, Dimitry Sibiryakov wrote:
09.12.2020 12:21, Alex Peshkoff via Firebird-devel wrote:
During such self assignment returned pointer can be invalidated by
setErrors() code before using input parameter which may cause crash
or wrong result.
Builtin implementation works fine
09.12.2020 12:41, Alex Peshkoff via Firebird-devel wrote:
setWarnings() does not use clear(), it's using save() internal call which does invalidates
memory in the very end and therefore is safe
Ah, I see. I looked at wrong save() overload.
--
WBR, SD.
Firebird-Devel mailing list, web int
On 12/9/20 2:31 PM, Dimitry Sibiryakov wrote:
09.12.2020 12:23, Alex Peshkoff via Firebird-devel wrote:
That depends upon getState() internals, is not it? :)
Yes. Firebird implementation recognizes zero error code in
getState() so I wonder if it is a specs' requirement or such
recognition
Hello All.
Remember me, please, when build metadata for a record should length value provided to
setLength() (and returned by getLength()) to be two bytes bigger when constructing VARCHAR
field?
--
WBR, SD.
Firebird-Devel mailing list, web interface at
https://lists.sourceforge.net/l
09.12.2020 18:31, Dimitry Sibiryakov wrote:
Remember me, please, when build metadata for a record should length value provided to
setLength() (and returned by getLength()) to be two bytes bigger when constructing VARCHAR
field?
Self-answer: no, get/setLength() work with exactly declared l
10 matches
Mail list logo
