Repository navigation
[Bug 63620] Fix GUI freeze when viewing response body with long lines - #694
Merged
Merged
Conversation
vlsi
force-pushed
the
split_text
branch
6 times, most recently
from
January 17, 2022 09:58
6af6f78 to
dd9e05f
Compare
Codecov Report
@@ Coverage Diff @@
## master #694 +/- ##
============================================
- Coverage 55.59% 55.54% -0.05%
- Complexity 10336 10345 +9
============================================
Files 1059 1061 +2
Lines 65053 65209 +156
Branches 7401 7432 +31
============================================
+ Hits 36163 36221 +58
- Misses 26340 26431 +91
- Partials 2550 2557 +7
Continue to review full report at Codecov.
|
vlsi
changed the title
vlsi
force-pushed
the
split_text
branch
2 times, most recently
from
January 17, 2022 10:26
34d45af to
17a050d
Compare
FSchumacher
reviewed
Jan 24, 2022
vlsi
force-pushed
the
split_text
branch
2 times, most recently
from
January 24, 2022 16:59
455a2cc to
cb8a120
Compare
Contributor
|
Looks really good to me. What keeps you from merging? |
Collaborator
Author
|
Thank you for the review and for yet another test case. my setup: macOS Monterey 12.1, Java 11.0.13. With Nimbus LaF it is really fast when browsing "view results tree". It is still usable, so we might want to merge the change and investigate the slowness with Darklaf later, however, I don't quite understand why Darklaf is slower since the thread spends time in It might be the font selected by Darklaf is somehow slower. Here's a stacktrace: /cc @weisJ |
Contributor
|
I suspect the performance hit is coming from font kerning. Try to set the font to |
Collaborator
Author
|
@weisJ , thanks for the hint. Initially, I thought the issue was ligature-related, however, ligatures are disabled by default. I completely forgot about the kerning. As I deactivated kerning the UI became much better, and the delays are like ~1sec when displaying 10MiB. On my machine, the kerning becomes noticeable at 9 and 10 scales (10-20KiB), so I am inclined to either completely disable the kerning for the results tab or automatically disable it after a certain threshold (e.g. after 10KiB). |
|
Shouldn't this setting be default in jmeter installation ?
…On Thu, Feb 3, 2022, 00:35 Jannis Weis ***@***.***> wrote:
I suspect the performance hit is coming from font kerning. Try to set the
font to comp.setFont(comp.getFont().derive(Collections.singletonMap(TextAttribute.KERNING,
0)))
—
Reply to this email directly, view it on GitHub
<#694 (comment)>, or
unsubscribe
<https://github.com/notifications/unsubscribe-auth/AAXQDSRBODMYVZTUPT46RPTUZGWR3ANCNFSM5MEDVDXQ>
.
You are receiving this because you are subscribed to this thread.Message
ID: ***@***.***>
|
Collaborator
Author
|
Which setting? |
|
comp.setFont(comp.getFont().derive(Collections.singletonMap(TextAttribute.KERNING,
0)))
Please see mail below.
…On Sat, Feb 5, 2022, 19:27 Vladimir Sitnikov ***@***.***> wrote:
Which setting?
—
Reply to this email directly, view it on GitHub
<#694 (comment)>, or
unsubscribe
<https://github.com/notifications/unsubscribe-auth/AAXQDSWVB4JFQCLFHVLNDR3UZVMXDANCNFSM5MEDVDXQ>
.
You are receiving this because you commented.Message ID:
***@***.***>
|
Collaborator
Author
|
The idea is to disable kerning for long texts by default |
|
Very good.
Thank you.
…On Sat, Feb 5, 2022, 19:39 Vladimir Sitnikov ***@***.***> wrote:
The idea is to disable kerning for long texts by default
—
Reply to this email directly, view it on GitHub
<#694 (comment)>, or
unsubscribe
<https://github.com/notifications/unsubscribe-auth/AAXQDSQRUXMDAWY7LBCWGZLUZVOD3ANCNFSM5MEDVDXQ>
.
You are receiving this because you commented.Message ID:
***@***.***>
|
Collaborator
Author
|
Just in case you wonder, here's a sample from macOS with hi-DPI screen. The first line is with kerning enabled, and the second one is with deactivated kerning.
|
Collaborator
Author
|
|
vlsi
force-pushed
the
split_text
branch
3 times, most recently
from
February 6, 2022 12:12
314b7a0 to
79fe7cd
Compare
Collaborator
Author
|
I've added a feature to automatically disable kerning when text length exceeds 10K, and now I can click through all the results. |
GUI gets unresponsive when the response contains long lines (e.g lines exceeding 100K chars).
The workaround is to add artificial line breaks.
We try to break on word boundaries first, and if not possible,
we break on char boundaries (e.g. large chunks of base64 data).
view.results.tree.max_line_size=110000 (defaults to 110K)
view.results.tree.soft_wrap_line_size=100000 (defaults to max_line_size/1.1)
CPU spin trace looks as follows:
"AWT-EventQue...
...
at sun.font.GlyphLayout.layout(Unknown Source)
...
at java.awt.font.TextLine.<init>(Unknown Source)
at java.awt.font.TextMeasurer.makeTextLineOnRange(Unknown Source)
at java.awt.font.TextMeasurer.getLayout(Unknown Source)
...
at javax.swing.text.JTextComponent.setDocument(Unknown Source)
at org.apache.jmeter.visualizers.SamplerResultTab.setTextOptimized(SamplerResultTab.java:695)
...
The treshold can be configured via text.kerning.max_document_size
|
Im using macos and 5.5 version - If i click on View Results tree when the request sent has body like 5k/10k records then the UI is getting frozen. Not sure if this fix covers this use case too ? @vlsi |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
6 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
GUI gets unresponsive when the response contains long lines (e.g lines exceeding 100K chars).
The workaround is to add artificial line breaks.
https://bz.apache.org/bugzilla/show_bug.cgi?id=63620
We try to break on word boundaries first, and if not possible,
we break on char boundaries (e.g. large chunks of base64 data).
For the reference, using zero-width space does not solve the problem, so I use
\nfor breaking lines.CPU spin trace looks as follows: