- 16 7月, 2016 7 次提交
-
-
由 Shreedhar Hardikar 提交于
-
由 Shreedhar Hardikar 提交于
-
由 Daniel Gustafsson 提交于
Since no callsite of these functions checks for a NULL return value and instead deref the returned pointer immediately we need to ensure that only valid values are ever returned. This also corrects a logic error in the enum range check which was introduced with the recent effort to silence compiler warnings.
-
由 Heikki Linnakangas 提交于
Reorder -lgpopt, -lnaucrates, -lgpdbcost, -lgpos, to make static compilation of ORCA to work. When linking with dynamic libraries, the order doesn't matter, but with static libraries, the options need to be given in dependency order. This doesn't change the way those libraries are actually linked. Whether we link with static or shared versions depends on which ones are found at link time.
-
由 Heikki Linnakangas 提交于
This gets rid of the libdxltranslators.so library. Instead, link the code in the gpopt directory straight into the backend binary. Much simpler. A pleasant side-effect of this is that the build can now also pick up the ORCA libraries from a path specified with "./configure --with-libs=<path>". It's no longer necessary to have OPTIMIZER and OBJDIR_DEFAULT Makefile variables set correctly, which is nice when building outside the gpAux/releng infrastructure. I'm hoping we can get rid of much of that infrastructure in the near future.
-
由 Heikki Linnakangas 提交于
The old mechanism extracted the version numbers from the Ivy config file, which doesn't do the right thing if you build without Ivy. Using the version headers is simpler, anyway.
-
由 Marbin Tan 提交于
The backup/restore context refactor had some regressions in netbackup and openbar backup and restore test suites. A summary of these fixes are: * Using the correct timestamp when performing an incremental backup * Generating timestamps in the correct order * Identifying whether compression is enabled in ddboost using the * correct file * Updating path to retreive file from on ddboost * Removing unused context variables * Removing dead code * Updates number/ordering in function argumenets Authors: Chris Hajas, Chumki Roy, Marbin Tan
-
- 15 7月, 2016 3 次提交
-
-
由 Adam Lee 提交于
1. speed up running by specifying proper files with sufficient size 2. add a pair of tests which connect to the same S3 file in parallel 3. revise sql scripts to make them neat Signed-off-by: NAdam Lee <ali@pivotal.io> Signed-off-by: NKuien Liu <kliu@pivotal.io>
-
由 Daniel Gustafsson 提交于
The 5.0.json file was added in commit 4665a8d5 which backported the variadic parameters patch from upstream but we are currently on 4.3 so remove awaiting the stamping of 5.0.
-
由 Daniel Gustafsson 提交于
Include all the supported formats for the FORMAT option on external tables in the errhint() call and perform minor copy editing to make it conform to the error message guidelines. Also remove the list of formats from the function comment since those lists are notorious for getting outdated (case in point).
-
- 14 7月, 2016 3 次提交
-
-
由 Jesse Zhang 提交于
By using authenticated access
-
由 Shreedhar Hardikar 提交于
-
由 Shreedhar Hardikar 提交于
-
- 13 7月, 2016 4 次提交
-
-
由 Adam Lee 提交于
Signed-off-by: NKuien Liu <kliu@pivotal.io>
-
由 Kuien Liu 提交于
Signed-off-by: NAdam Lee <ali@pivotal.io>
-
由 Kenan Yao 提交于
dispatcher
-
由 foyzur 提交于
* Preventing multiple ResLockPortal calls for the same portal when running multiple queries via PortalRunMulti by correctly populating canSetTag in PlannedStmt from Query object during DXL to PlannedStmt translation. * ICG tests for checking if ORCA correctly populates canSetTag.
-
- 12 7月, 2016 7 次提交
-
-
由 Kuien Liu 提交于
Signed-off-by: NAdam Lee <ali@pivotal.io>
-
由 Lucas Lonergan 提交于
Added a link to a quickstart docker guide located on the Pivotal Engineering Journal.
-
由 Shreedhar Hardikar 提交于
-
由 Shreedhar Hardikar 提交于
-
由 Shreedhar Hardikar 提交于
ExprTreeGeneratorInfo object keeps track of metadata information needed to generate code for expression trees. Refactor ExprTreeGenerators to always use econtext->ecxt_scantuple to generate the functions. We also have an extra slot_check block to fall back if it is called with a different slot - which shouldn't happen for Agg and Scan nodes.
-
由 Shreedhar Hardikar 提交于
Split ExecVariableListCodegen::GenerateExecVariableList into 2 parts, one for managing the generation of ExecVariableList itself, and another for of slot_getattr(). Also move the implementation of the code The SlotGetAttr generator uses a static method called from both ExecVariableList and ExecEvalExpr to generate appropriate slot_getattr(). The actual generation is wrapped to handle the case when the generation fails, since there is no infrastructure right now to implement shared code generation classes that automatically clean up created functions.
-
由 Nikos Armenatzoglou 提交于
-
- 11 7月, 2016 13 次提交
-
-
由 Peter Eisentraut 提交于
The code was assigning a (Datum) 0 to a void pointer. That creates a warning from clang 3.4. It was probably a thinko to begin with.
-
由 Daniel Gustafsson 提交于
Shifting a negative signed value is undefined behavior in C, instead replace with the defined constant that was intended to reach with the bitshift. It's not entirely clear to me *why* setting such a low value is of interest given that further operations on it can overflow to being positive but this keeps the current functionality with one less compiler warning.
-
由 Daniel Gustafsson 提交于
Normally the compiler can figure out when logical chains have inverted truths like: if (foo) {} else if (!foo) {}. When there are more variables involved this doesn't always work and we get compiler warnings for unset variables which are indeed set. Remove and rewrite offending logic to be clearer for the compiler and the human reader.
-
由 Daniel Gustafsson 提交于
Dereferencing address zero is undefined behavior in C and the compiler is free to optimize any such calls away (even though that is rare). Ensure that any optimizations are avoided by declaring the variable volatile and at the same time get rid of a clang compiler warning.
-
由 Michael Meskes 提交于
-
由 Daniel Gustafsson 提交于
Remove a pointless redefinition of the prefixsel variable and move the selec setting into under the else clause as it otherwise over- writes the value we just calculated. This shortcircuited the GPDB code and fell back on the default upstream behaviour so the effect was nullified.
-
由 Daniel Gustafsson 提交于
These variables are unitialized but will be set before used, clang doesn't always figure this out though so give a helping hand and set a value at initialization. The initalizations in Gin scanning were later performed upstream in 87b8db37 as part of a larger body of work.
-
由 Daniel Gustafsson 提交于
Since the actual datatype used for an enum is implementation specific per each compiler in both C90 and C99 we can't rely on it automatically being an unsigned integer even though it most likely will be in this case. Inverting the check and testing for it's negation will achieve the range check that was intended without compiler warnings issued for comparison of unsigned enum (-Wtautological-compare in clang). Adding a comment explaining this both in the code and the headerfile. Note that the existing coding was actually incorrect as it was testing for the tabletype ranges with logical AND and not logical OR which would've been correct, the severity of this bug must be considered quite low though.
-
由 Tom Lane 提交于
locale-dependent character classification properly when the database encoding is UTF8. The previous coding worked okay in single-byte encodings, or in any case for ASCII characters, but failed entirely on multibyte characters. The fix assumes that the <wctype.h> functions use Unicode code points as the wchar representation for Unicode, ie, wchar matches pg_wchar. This is only a partial solution, since we're still stupid about non-ASCII characters in multibyte encodings other than UTF8. The practical effect of that is limited, however, since those cases are generally Far Eastern glyphs for which concepts like case-folding don't apply anyway. Certainly all or nearly all of the field reports of problems have been about UTF8. A more general solution would require switching to the platform's wchar representation for all regex operations; which is possible but would have substantial disadvantages. Let's try this and see if it's sufficient in practice.
-
由 Daniel Gustafsson 提交于
The BM_MAX_TUPLES_PER_PAGE macro will evaluate to the uint16 max so asserting that an OffsetNumber is <= is always true given the typedef on OffsetNumber from uint16. This fixes -Wtautological-constant-out-of-range-compare warnings.
-
由 Daniel Gustafsson 提交于
The @outarr array may at this point be empty and using subscript -1 to reach the last element will thus throw an uninitialized error. Test for array contents before accessing any elements to avoid the error and to exit early since the block by definition isn't of interest. Observed in the wild for the bfv_partition test suite.
-
由 Daniel Gustafsson 提交于
This is by no means all the offending lines in this file, but they were the ones I had to read in order to figure out a bug. Remove the worst offenders such as three-space indentation, commented out code and and blocks at the wrong level. A larger cleanup is at some point needed but piece by piece we are getting this code into shape.
-
由 Adam Lee 提交于
Signed-off-by: NAdam Lee <ali@pivotal.io>
-
- 09 7月, 2016 3 次提交
-
-
由 kaknikhil 提交于
-
由 Shoaib Lari 提交于
Make the error message more descriptive when cannot enlarge readRecordBuf.
-
由 Daniel Gustafsson 提交于
When the workfile caching support was removed the number of columns in the gp_workfile_mgr_cache_entries view was decreased to twelve. Fix the array offsets to avoid writing past the end of the array as they were accidentally left out in that commit. Also move the assertion for view width to before the use of the tupledesc's in question as it's more useful there than after we've accessed what we're asserting.
-