Tags: sofong5/hhvm
Tags
Updating submodules Summary: GitHub commits: facebook/CacheLib@4845975 facebook/fbthrift@9bcd08f facebook/folly@ab0af93 facebookexperimental/edencommon@37ceac1 facebookincubator/fizz@f4d134c facebookincubator/velox@11c216d Reviewed By: jurajh-fb fbshipit-source-id: 3af00e08b0705201d794d636b32805b510fa9258
(2/2) Remove remote file store main logic Summary: The stack of diffs ending in D38663620 (facebook@fcda049) introduced a hh command line to generate an experimental remote file store for sub1m. This has not been in use and the manifold bucket hack_sub1m_file_store is going away. This is the second diff that removes the main logic to fetch and store all WWW file content. Reviewed By: bobrenjc93 Differential Revision: D40913283 fbshipit-source-id: 65ab45a3b8cbdb6a1fec9497abe997924a056ead
Updating submodules Summary: GitHub commits: facebook/fbthrift@7de8d41 facebook/rocksdb@9f3475e facebookincubator/velox@dc7809f Reviewed By: yns88 fbshipit-source-id: 11d1fbb014cb7ae97874ea119f8b9eb03353489f
Fix control pass
Summary:
The control pass wasn't correctly collapsing the entry block for a function with a default param:
```
<<__Memoize>>
function f(int $x, int $incr = 1): int {
return $x;
}
```
it was generating (note the empty `b0`):
```
function f($x, $incr @ b1 ("1")) {
b0:
jmp to b2
b1:
...
enter to b2
b2:
...
memo_get [0, 1] to b3 else b4
b3(%10):
ret %10
b4:
...
ret %15
}
```
instead of the expected:
```
function f($x, $incr @ b3 ("1")) {
b0:
...
memo_get [0, 1] to b1 else b2
b1(%6):
ret %6
b2:
...
ret %11
b3:
...
enter to b0
}
```
the problem was that the code trying to avoid creating a critical edge was being too aggressive.
Fixed it and added a bunch of unit tests to verify the behavior (the example above is `test8`).
Reviewed By: lavenzg
Differential Revision: D40790269
fbshipit-source-id: 5e9afb89f22c06f744ceb0a2c8d80a638b850e60
Eliminate builtinType field from ParamInfo Reviewed By: ricklavoie Differential Revision: D40812025 LaMa Project: L1122815 fbshipit-source-id: 0f0c65188d24cdb27fc17e00a5f1a837f0c5cf6a
Updating submodules Summary: GitHub commits: facebook/CacheLib@d96eb65 facebook/fbthrift@a839ddc facebook/mcrouter@00a4130 facebook/rocksdb@9079895 facebookexperimental/edencommon@b522217 facebookexperimental/rust-shed@dab13dd facebookincubator/fizz@165e5a7 facebookincubator/katran@0249eab facebook/mvfst@60a67fd facebookincubator/velox@63086a0 facebookresearch/beanmachine@d222e69 Reviewed By: yns88 fbshipit-source-id: 0796a24891415ef719da32f4241687bf17a66ef6
Fix build for 32-bit pointers Summary: Removes the assumption that `uintptr_t` is 64 bits. Instead of hardcoding 52 bits for local IDs, the allocation is now 52 bits when `uintptr_t` is 64 bits and 26 bits when `uintptr_t` is 32 bits. Reviewed By: roddym Differential Revision: D40687166 fbshipit-source-id: 3eb9cf9a3a99ffb74f4871504d73fcecc26c91a2
make array_rand, fb_id_to_db, DBSpec::getDbspecForDbid, DBSpec::getIs… …GeneralTest leak_safe (HHI) Summary: For D35405063 (facebook@8d7354b). Reviewed By: oulgen Differential Revision: D40734612 fbshipit-source-id: 64e4af5196392c2dc1d04259bb35385582ac735e
update XLOG() to strip buck2 header prefixes Summary: Buck v2 uses a different directory layout for its header directories than Buck v1. Update the xlog code with knowledge of how to parse and strip the buck2 paths. This also updates the code to use preprocessor macros to decide at build time whether we need to support buck1 or buck2 style paths. For CMake-based builds we do not attempt to look for either. In theory we could need to support multiple build systems if we ever linked together libraries built with multiple different build systems. Reviewed By: Gownta Differential Revision: D40403871 fbshipit-source-id: 4f5b7aa9aec1adf4e3535efaea203268e8d20034
PreviousNext