programmingk...@gmail.com[*] reported a runtime failure on a
32-bit Mac OS compilation, where "%"PRId64 expands to "%qd".
Fortunately, we had very few spots that were relying on our
pseudo-printf JSON parsing of 64-bit numbers, so it was
easier to just convert callers and rip out the 64-bit support.
The remaining uses of pseudo-printf handling are more complex;
there are only 3 users in the released codebased, but LOTS of
users in the testsuite (via wrapper functions like qmp()); I
will be posting a followup series that rips out the remaining
uses of dynamic JSON, but it will be 2.9 material, while
these three patches qualify for 2.8.
[*] git log shows the name John, but the particular email that
sparked this only stated the non-descript name 'G 3', which
makes it a bit hard for me to know which form is preferred
when lending credit.
In patch 3, I chose to remove existing support for 64-bit
dynamic JSON on all platforms, so that we quickly see a failure
in an attempt to use such a construct even on mainstream glibc
development environments, rather than only on the less-frequent
32-bit Mac OS build environment.
Eric Blake (3):
qmp-event: Avoid qobject_from_jsonf("%"PRId64)
test-qga: Avoid qobject_from_jsonf("%"PRId64)
qapi: Drop support for qobject_from_jsonf("%"PRId64)
qapi/qmp-event.c | 12 +++++++-----
qobject/json-lexer.c | 28 ----------------------------
qobject/json-parser.c | 5 -----
tests/check-qjson.c | 10 ----------
tests/test-qga.c | 9 +++++++--
tests/test-qobject-input-visitor.c | 5 +++--
6 files changed, 17 insertions(+), 52 deletions(-)
--
2.7.4
