MacPython linux build failing #34114
Comments
TomAugspurger
added
Bug
Needs Triage
|
These are the commits to pandas from the 1st to the 4th |
|
@jbrockmendel or @WillAyd do you haev any guesses on this? Looking at @cython.cdivision
cdef int pqyear(int64_t ordinal, int freq):
cdef:
int year, quarter
get_yq(ordinal, freq, &quarter, &year)
return yearDo you know where |
|
get_yq calls I'll look at this more closely after some caffeine |
|
Do you know of any reason why the lhs of those assign to array indices? I think more correct to assign to a dereferenced pointer which might get rid of the warning.
…Sent from my iPhone
On May 12, 2020, at 7:31 AM, jbrockmendel ***@***.***> wrote:
get_yq calls quarter[0] = DtoQ_yq(unix_date, &af_info, year) and in DtoQ_yq year is set with year[0] = dts.year
I'll look at this more closely after some caffeine
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub, or unsubscribe.
|
|
Nvm that’s a Cython requirement.
https://cython.readthedocs.io/en/latest/src/userguide/language_basics.html#types
So probably fine
…Sent from my iPhone
On May 12, 2020, at 8:23 AM, William Ayd ***@***.***> wrote:
Do you know of any reason why the lhs of those assign to array indices? I think more correct to assign to a dereferenced pointer which might get rid of the warning.
Sent from my iPhone
>> On May 12, 2020, at 7:31 AM, jbrockmendel ***@***.***> wrote:
>>
>
> get_yq calls quarter[0] = DtoQ_yq(unix_date, &af_info, year) and in DtoQ_yq year is set with year[0] = dts.year
>
> I'll look at this more closely after some caffeine
>
> —
> You are receiving this because you were mentioned.
> Reply to this email directly, view it on GitHub, or unsubscribe.
|
Closed
TomAugspurger
added
Build
|
Why does |
Best guess is thats leftover from a time when division was actually done within pqyear |
|
Cannot this cause this problem, I mean? |
I don't see how, but you're welcome to try removing it and see if it helps. |
|
Have no clue in the meantime how to do it with all those azure links at the top, sorry... Just dropped the idea... |
|
FWIW this is reproducible in the manylinux1_x86_64 docker file. FROM quay.io/pypa/manylinux1_x86_64:latest
RUN git clone --depth=1 https://github.com/pandas-dev/pandas \
&& cd pandas \
&& /opt/python/cp38-cp38/bin/python -m pip install numpy cython python-dateutil pytz
WORKDIR pandas
RUN /opt/python/cp38-cp38/bin/python setup.py build_ext -iI don't really see any differences in the generated C code. I'll try to bisect the changes to see which commit this started failing on. |
Closed
|
Pretty obvious in retrospect: aee8e11 is the first "bad" commit, which elevated warnings to errors. That doesn't help us identify why it's only failing with this version of gcc / whatever else matters. |
|
There are some other warnings |
|
Do you know what Cython version that is using? What's strange to me is that taking say
|
|
@TomAugspurger do you know how to get the generated hashtable.c file out of the container? Would be helpful to review that alone |
TomAugspurger
added a commit
to TomAugspurger/pandas
that referenced
this issue
May 27, 2020
TomAugspurger
added a commit
that referenced
this issue
May 27, 2020
|
Repurposing this issue for the 32-bit specific build issues now. I've updated the original post. @WillAyd I think you're correct in #34114 (comment). This is from the and if k != table.n_bucketsI've uploaded the generated hashtable.c file to https://gist.github.com/TomAugspurger/1f60b2059a103e6d36c02a0d07200e59 |
|
In So In So is |
|
May have fixed the hashtable issues by defining IIUC, we have int _tokenize_helper(parser_t *self, size_t nrows, int all) {
...
int64_t start_lines = self->lines;
but typedef struct parser_t {
...
uint64_t lines; // Number of (good) lines observed
... |
|
Nice! I think that change is consistent with other areas in hashing as well.
For the parsing issue *might* be easiest to just make that signed
…Sent from my iPhone
On May 27, 2020, at 11:53 AM, Tom Augspurger ***@***.***> wrote:
May have fixed the hashtable issues by defining k as khiter_t. Some more failures after that.
gcc -pthread -Wno-unused-result -Wsign-compare -DNDEBUG -g -fwrapv -O3 -Wall -fPIC -DNPY_NO_DEPRECATED_API=0 -Ipandas/_libs -I./pandas/_libs/tslibs -I./pandas/_libs -Ipandas/_libs/src/klib -I/opt/python/cp38-cp38/lib/python3.8/site-packages/numpy/core/include -I/opt/python/cp38-cp38/include/python3.8 -c pandas/_libs/src/parser/tokenizer.c -o build/temp.linux-i686-3.8/pandas/_libs/src/parser/tokenizer.o -Werror
pandas/_libs/src/parser/tokenizer.c: In function ‘tokenize_bytes’:
pandas/_libs/src/parser/tokenizer.c:629:39: error: comparison between signed and unsigned integer expressions [-Werror=sign-compare]
if (line_limit > 0 && self->lines == start_lines + line_limit) { \
^
pandas/_libs/src/parser/tokenizer.c:648:20: note: in expansion of macro ‘END_LINE_STATE’
#define END_LINE() END_LINE_STATE(START_RECORD)
^
pandas/_libs/src/parser/tokenizer.c:760:21: note: in expansion of macro ‘END_LINE’
END_LINE();
^
pandas/_libs/src/parser/tokenizer.c:629:39: error: comparison between signed and unsigned integer expressions [-Werror=sign-compare]
if (line_limit > 0 && self->lines == start_lines + line_limit) { \
^
pandas/_libs/src/parser/tokenizer.c:648:20: note: in expansion of macro ‘END_LINE_STATE’
#define END_LINE() END_LINE_STATE(START_RECORD)
^
pandas/_libs/src/parser/tokenizer.c:775:21: note: in expansion of macro ‘END_LINE’
END_LINE();
^
pandas/_libs/src/parser/tokenizer.c:629:39: error: comparison between signed and unsigned integer expressions [-Werror=sign-compare]
if (line_limit > 0 && self->lines == start_lines + line_limit) { \
^
pandas/_libs/src/parser/tokenizer.c:648:20: note: in expansion of macro ‘END_LINE_STATE’
#define END_LINE() END_LINE_STATE(START_RECORD)
^
pandas/_libs/src/parser/tokenizer.c:798:21: note: in expansion of macro ‘END_LINE’
END_LINE();
^
pandas/_libs/src/parser/tokenizer.c:629:39: error: comparison between signed and unsigned integer expressions [-Werror=sign-compare]
if (line_limit > 0 && self->lines == start_lines + line_limit) { \
^
pandas/_libs/src/parser/tokenizer.c:648:20: note: in expansion of macro ‘END_LINE_STATE’
#define END_LINE() END_LINE_STATE(START_RECORD)
^
pandas/_libs/src/parser/tokenizer.c:840:21: note: in expansion of macro ‘END_LINE’
END_LINE();
^
pandas/_libs/src/parser/tokenizer.c:629:39: error: comparison between signed and unsigned integer expressions [-Werror=sign-compare]
if (line_limit > 0 && self->lines == start_lines + line_limit) { \
^
pandas/_libs/src/parser/tokenizer.c:648:20: note: in expansion of macro ‘END_LINE_STATE’
#define END_LINE() END_LINE_STATE(START_RECORD)
^
pandas/_libs/src/parser/tokenizer.c:870:29: note: in expansion of macro ‘END_LINE’
END_LINE();
^
pandas/_libs/src/parser/tokenizer.c:629:39: error: comparison between signed and unsigned integer expressions [-Werror=sign-compare]
if (line_limit > 0 && self->lines == start_lines + line_limit) { \
^
pandas/_libs/src/parser/tokenizer.c:648:20: note: in expansion of macro ‘END_LINE_STATE’
#define END_LINE() END_LINE_STATE(START_RECORD)
^
pandas/_libs/src/parser/tokenizer.c:879:25: note: in expansion of macro ‘END_LINE’
END_LINE();
^
pandas/_libs/src/parser/tokenizer.c:629:39: error: comparison between signed and unsigned integer expressions [-Werror=sign-compare]
if (line_limit > 0 && self->lines == start_lines + line_limit) { \
^
pandas/_libs/src/parser/tokenizer.c:648:20: note: in expansion of macro ‘END_LINE_STATE’
#define END_LINE() END_LINE_STATE(START_RECORD)
^
pandas/_libs/src/parser/tokenizer.c:916:21: note: in expansion of macro ‘END_LINE’
END_LINE();
^
pandas/_libs/src/parser/tokenizer.c:629:39: error: comparison between signed and unsigned integer expressions [-Werror=sign-compare]
if (line_limit > 0 && self->lines == start_lines + line_limit) { \
^
pandas/_libs/src/parser/tokenizer.c:648:20: note: in expansion of macro ‘END_LINE_STATE’
#define END_LINE() END_LINE_STATE(START_RECORD)
^
pandas/_libs/src/parser/tokenizer.c:964:21: note: in expansion of macro ‘END_LINE’
END_LINE();
^
pandas/_libs/src/parser/tokenizer.c:629:39: error: comparison between signed and unsigned integer expressions [-Werror=sign-compare]
if (line_limit > 0 && self->lines == start_lines + line_limit) { \
^
pandas/_libs/src/parser/tokenizer.c:648:20: note: in expansion of macro ‘END_LINE_STATE’
#define END_LINE() END_LINE_STATE(START_RECORD)
^
pandas/_libs/src/parser/tokenizer.c:1031:21: note: in expansion of macro ‘END_LINE’
END_LINE();
^
pandas/_libs/src/parser/tokenizer.c:629:39: error: comparison between signed and unsigned integer expressions [-Werror=sign-compare]
if (line_limit > 0 && self->lines == start_lines + line_limit) { \
^
pandas/_libs/src/parser/tokenizer.c:648:20: note: in expansion of macro ‘END_LINE_STATE’
#define END_LINE() END_LINE_STATE(START_RECORD)
^
pandas/_libs/src/parser/tokenizer.c:1049:21: note: in expansion of macro ‘END_LINE’
END_LINE();
^
pandas/_libs/src/parser/tokenizer.c:629:39: error: comparison between signed and unsigned integer expressions [-Werror=sign-compare]
if (line_limit > 0 && self->lines == start_lines + line_limit) { \
^
pandas/_libs/src/parser/tokenizer.c:648:20: note: in expansion of macro ‘END_LINE_STATE’
#define END_LINE() END_LINE_STATE(START_RECORD)
^
pandas/_libs/src/parser/tokenizer.c:1059:21: note: in expansion of macro ‘END_LINE’
END_LINE();
^
pandas/_libs/src/parser/tokenizer.c:629:39: error: comparison between signed and unsigned integer expressions [-Werror=sign-compare]
if (line_limit > 0 && self->lines == start_lines + line_limit) { \
^
pandas/_libs/src/parser/tokenizer.c:1062:25: note: in expansion of macro ‘END_LINE_STATE’
END_LINE_STATE(EAT_WHITESPACE);
^
pandas/_libs/src/parser/tokenizer.c:644:39: error: comparison between signed and unsigned integer expressions [-Werror=sign-compare]
if (line_limit > 0 && self->lines == start_lines + line_limit) { \
^
pandas/_libs/src/parser/tokenizer.c:1065:25: note: in expansion of macro ‘END_LINE_AND_FIELD_STATE’
END_LINE_AND_FIELD_STATE(START_FIELD);
^
pandas/_libs/src/parser/tokenizer.c:629:39: error: comparison between signed and unsigned integer expressions [-Werror=sign-compare]
if (line_limit > 0 && self->lines == start_lines + line_limit) { \
^
pandas/_libs/src/parser/tokenizer.c:1076:25: note: in expansion of macro ‘END_LINE_STATE’
END_LINE_STATE(START_RECORD);
^
pandas/_libs/src/parser/tokenizer.c:1093:41: error: comparison between signed and unsigned integer expressions [-Werror=sign-compare]
self->lines == start_lines + line_limit) {
^
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub, or unsubscribe.
|
|
Another one :) |
|
Guessing from phone but if those are compared to unsigned integers than probably want to add a U or UL Suffix to the constants
…Sent from my iPhone
On May 27, 2020, at 12:12 PM, Tom Augspurger ***@***.***> wrote:
Another one :)
pandas/_libs/tslibs/period.c:674:9: error: this decimal constant is unsigned only in ISO C90 [-Werror]
{0, 1, 60, 3600, 3600000, 3600000000, 3600000000000},
^
pandas/_libs/tslibs/period.c:1358:3: error: this decimal constant is unsigned only in ISO C90 [-Werror]
__pyx_e_6pandas_5_libs_6tslibs_6period_INT32_MIN = -2147483648L
^
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub, or unsubscribe.
|
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
There are build failures with the manylinux1 32-bit docker image.
To reproduce
I've uploaded the generated hashtable.c file here.
#34409 fixed the following issues.
https://dev.azure.com/pandas-dev/pandas-wheels/_build/results?buildId=35250&view=logs&j=e9afaa34-1a0f-534e-78fc-fae528ccd915&t=4685e914-c192-594f-c25f-e6cef98b7ba4
https://dev.azure.com/pandas-dev/pandas-wheels/_build/results?buildId=34768&view=results from May 3rd may be the first one with this warning. https://dev.azure.com/pandas-dev/pandas-wheels/_build/results?buildId=34727&view=results from May 2nd built successfully.
The text was updated successfully, but these errors were encountered: