Don't include any captured stdout/stderr/log output to the report

[bigunyak]

Hi,
To reduce report file size, I'm trying to exclude any captured stdout/stderr/log output.
I'm running tests with the following command:

pytest --html=report.html --self-contained-html --capture=no --show-capture=no tests/

With this I still get the Captured log call section for every test no matter if it pass or not.
So the question, how to not include any captured output to the report?

[bigunyak]

The --show-capture option seems to have absolutely no effect on what gets included into the report.
Could you please help to understand how capturing suppose to work?

[davehunt]

--show-capture is not related to the HTML report, but I like the idea. I'll label this as an enhancement and maybe somebody would be interested in working on it. I don't think it would be too hard, we'd just need to get the value from the configuration and make the writing to the report conditional based on it.

[adriangb]

This honestly seems like a bug to me, not an enhancement. --capture=no should disable all capture.

[BeyondEvil]

I agree with @adriangb.

What do you think @davehunt ?

[rochacbruno]

what I understood is that --capture is a pytest argument not a pytest-html argument.

So the enhancement would be adding the feature to read that arguments and do something.

[BeyondEvil]

I think it makes sense for the plugin to ”obey” the flag.

…

On 5 Sep 2019, 12:52 +0200, Bruno Rocha ***@***.***>, wrote: what I understood is that --capture is a pytest argument not a pytest-html argument. So the enhancement would be adding the feature to read that arguments and do something. — You are receiving this because you commented. Reply to this email directly, view it on GitHub, or mute the thread.

[adriangb]

what I understood is that --capture is a pytest argument not a pytest-html argument.

So the enhancement would be adding the feature to read that arguments and do something.

Yes, --capture and --show-capture are Pytest options, but regardless it would be nice to have these options in the plugin (there is a reason they exist in Pytest), so why not make the plugin follow whatever Pytest does? Then to go the extra mile --html-report-capture and --html-report-show-capture could be added which override --capture and --show-capture for the HTML report.

In terms of implementation, I am really very new to Pytest so take this with a grain of salt, but I noticed the same thing (capturing of log output despite use of --capture=no) happens with pytest-json and Junit XML reports, which leads me to thing that Pytest itself is not completely respecting the flag (if it was, these plugins should never be able to get the log output in the first place)

[davehunt]

I still think this is an enhancement, and a worthwhile one. Perhaps pytest should be more clear that the capture arguments only relate to the console logging, and that plugins are not required to respect them. If you disagree then I would suggest opening a bug against pytest that prevents plugins from having access to this content if the arguments are used.

Either way, I don't think we want to duplicate these arguments for each plugin.

[sjkcarnahan]

I would like to tack on that this would be a great option.

[hello-ming]

Any progress here? It's better if we could control disabling stdout or stderr or all.

Wondering why captured stderr/log call contents are the same in my html report, I'm using xdist plugin

[kstsai]

FYI,

I used
pytest --html=1.htm --self-contained-html --capture=tee-sys

the generated 1.thm
DOES include captured stdout output nearby relevant test case.

[longelier]

How I use stdout, is to display comments/messages as each test case progresses, and it shows up in the html report. So please don't change that.
I have multiple variables to check in each testcase and I want to show in the report what the value is at that point, before moving on. Call these checkpoints, and there are many before it gets to the final checkpoint...the completion of each test case. These extra messages tell me, and others reader, what the test case did and what values at that time of the message. Also some confidence that it ran as reader expected. Seeing just pass/fail on a report is not as helpful, but seeing it with some extra proof that it really did pass is.

Example output from html report, that each test case spits out:
-----------------------------Captured stdout setup------------------------------
(presetup) Starting.
------------------------------Captured stdout call------------------------------
(test_case1) Starting.
PASS1: Apple color matches 'red'.
FAIL2: Apple size does not match 'big'.
------------------------------Captured stderr call------------------------------
20200612 060306 INFO: (test_case1) Apple=(green)
-------------------------------Captured log call--------------------------------
[32mINFO [0m test_case1:test_case1.py:45 (test_case1) Apple=(green)

I like the idea of being able to disable/enable some or all of the capture.
In above 'Captured log call' section, I specifically used logging.info("test_case1) Apple=(green)") to display a message...and when I ran it at the command prompt, there were unexpected ASCII characters(that I don't want). It would be nice to disable this part of the capture.
Running within PyCharm doesn't show the ASCII, but I need it run at command line more than an IDE.

[yogeshwar1996]

Its anoying to have logs captured in html report. The file size becomes >1GB. Kindly add the feature to disable log capturing in report

[longelier]

Agreed. There should be a away disable each of out the outputs, so the html report does not include it...thereby keeping the file size smaller.
The output in my html report, is from my communicating with network switches, and can really accumulate in each of the below sections:

1. Captured stdout setup
2. Captured stderr setup
3. Captured log setup
4. Captured stdout call
5. Captured stderr call
6. Captured log call

[NicolasAgra]

I just have found this issue and I am having a similar problem but maybe I should create another one. Let me know @davehunt . I created a logger to see in live the logs through the normal stdout stream my customs messages and avoid dependency logs. I usually run in the terminal individual tests with the "-s" option to see logs in live. However, when I do this with the --html option, the report generates just fine but I get all duplicated when showing:

 ------------------------------Captured stdout call------------------------------ 
21:16:20.224 - INFO - 

TEST DESCRIPTION: Some description
21:16:20.849 - INFO - 🏃 MSJ: {"error":"invalid_token"}

 -------------------------------Captured log call-------------------------------- 
INFO     mylib.py:36 

TEST DESCRIPTION: Some description
INFO     mylib.py:24 MSJ: {"error":"invalid_token"}

Is there a way to avoid log call? Since the option "--no-print-logs" was deprecated from older versions of pytest, the option "--show-capture-logs=no" does not apply in the same way. I have the latest versions:

pytest==6.1.1
pytest-csv==2.0.2
pytest-forked==1.3.0
pytest-html==2.1.1
pytest-metadata==1.10.0
pytest-reportportal==5.0.5
pytest-sugar==0.9.4
pytest-xdist==2.1.0

My log definition:

import sys
import logging

# Logging stuff
logger = logging.getLogger(__name__)
logger.setLevel(logging.INFO)

# Define messages format
formatter = logging.Formatter('%(asctime)s.%(msecs)03d - %(levelname)s - %(message)s', datefmt='%H:%M:%S')

# Define handler
handler = logging.StreamHandler(sys.stdout)
handler.setFormatter(formatter)
logger.addHandler(handler)

Regards!

[BeyondEvil]

This has been a long running issue. Would @gnikonorov or @ssbarnea have some time to just take a quick look to see if there's an easy/fast fix for this?

[gnikonorov]

This has been a long running issue. Would @gnikonorov or @ssbarnea have some time to just take a quick look to see if there's an easy/fast fix for this?

I can take a look @BeyondEvil!

[swbchangle]

should_skip_captured_output = pytest_capture_value == "no"
if report.outcome == "failed" and not should_skip_captured_output:
    should_skip_captured_output = pytest_show_capture_value == "no"
if not should_skip_captured_output:
    self._populate_html_log_div(log, report)

From this code I understand that
If --capture is not "no", pytest-html puts it in html unless --show-capture=no is specified.
If --capture=no , pytest-html ignores anything in stdout/stderr/log, regardless of --show-capture=no specified or not.

But --capture only controls stdout/stderr. I am using logging module to log messages. I set --capture=no. In this case, nothing will be logged in html.

[mubarakl-netapp]

I am still facing thi

I just have found this issue and I am having a similar problem but maybe I should create another one. Let me know @davehunt . I created a logger to see in live the logs through the normal stdout stream my customs messages and avoid dependency logs. I usually run in the terminal individual tests with the "-s" option to see logs in live. However, when I do this with the --html option, the report generates just fine but I get all duplicated when showing:

 ------------------------------Captured stdout call------------------------------ 
21:16:20.224 - INFO - 

TEST DESCRIPTION: Some description
21:16:20.849 - INFO - 🏃 MSJ: {"error":"invalid_token"}

 -------------------------------Captured log call-------------------------------- 
INFO     mylib.py:36 

TEST DESCRIPTION: Some description
INFO     mylib.py:24 MSJ: {"error":"invalid_token"}

Is there a way to avoid log call? Since the option "--no-print-logs" was deprecated from older versions of pytest, the option "--show-capture-logs=no" does not apply in the same way. I have the latest versions:

pytest==6.1.1
pytest-csv==2.0.2
pytest-forked==1.3.0
pytest-html==2.1.1
pytest-metadata==1.10.0
pytest-reportportal==5.0.5
pytest-sugar==0.9.4
pytest-xdist==2.1.0

My log definition:

import sys
import logging

# Logging stuff
logger = logging.getLogger(__name__)
logger.setLevel(logging.INFO)

# Define messages format
formatter = logging.Formatter('%(asctime)s.%(msecs)03d - %(levelname)s - %(message)s', datefmt='%H:%M:%S')

# Define handler
handler = logging.StreamHandler(sys.stdout)
handler.setFormatter(formatter)
logger.addHandler(handler)

Regards!

I am also facing this issue. Is there any fix or workaround? Please help

[QiaoqinXiao]

I just have found this issue and I am having a similar problem but maybe I should create another one. Let me know @davehunt . I created a logger to see in live the logs through the normal stdout stream my customs messages and avoid dependency logs. I usually run in the terminal individual tests with the "-s" option to see logs in live. However, when I do this with the --html option, the report generates just fine but I get all duplicated when showing:

 ------------------------------Captured stdout call------------------------------ 
21:16:20.224 - INFO - 

TEST DESCRIPTION: Some description
21:16:20.849 - INFO - 🏃 MSJ: {"error":"invalid_token"}

 -------------------------------Captured log call-------------------------------- 
INFO     mylib.py:36 

TEST DESCRIPTION: Some description
INFO     mylib.py:24 MSJ: {"error":"invalid_token"}

Is there a way to avoid log call? Since the option "--no-print-logs" was deprecated from older versions of pytest, the option "--show-capture-logs=no" does not apply in the same way. I have the latest versions:

pytest==6.1.1
pytest-csv==2.0.2
pytest-forked==1.3.0
pytest-html==2.1.1
pytest-metadata==1.10.0
pytest-reportportal==5.0.5
pytest-sugar==0.9.4
pytest-xdist==2.1.0

My log definition:

import sys
import logging

# Logging stuff
logger = logging.getLogger(__name__)
logger.setLevel(logging.INFO)

# Define messages format
formatter = logging.Formatter('%(asctime)s.%(msecs)03d - %(levelname)s - %(message)s', datefmt='%H:%M:%S')

# Define handler
handler = logging.StreamHandler(sys.stdout)
handler.setFormatter(formatter)
logger.addHandler(handler)

Regards!

Do you have any solutions? I am also facing this issue, because the duplicated logs, it has difficult to check the failures via log.

[BeyondEvil]

@QiaoqinXiao

try 4.0.0rc4