search

jromang / picochess

GNU General Public License v3.0
74 stars 54 forks source link

Picochess stalls on startup #94

Closed ifekali closed 9 years ago

ifekali commented 9 years ago

On startup Picochess will not respond to first move, it stalls for a couple of minutes before responding. Meanwhile selecting the levels, time controls and books DOES work, it responds normally. See video:

https://youtu.be/MQzuEi4U8-8

I'm on master branch in the video. (So it's not updating, right?)

djdekker commented 9 years ago

Happens here, too. Updating to 0.41 solved the issue.

ifekali commented 9 years ago

Nope, updating doesn't solve it.

ifekali commented 9 years ago

BTW, this bug was not introduced with 0.41, I noticed it months ago ...

sshivaji commented 9 years ago

I noticed this issue. Will solve this and a few other stability issues this weekend. Can you reproduce this issue if you change logging to info level?

On Jun 12, 2015, at 3:07 PM, Izmet Fekali notifications@github.com wrote:

BTW, this bug was not introduced with 0.41, I noticed it months ago ...

— Reply to this email directly or view it on GitHub https://github.com/jromang/picochess/issues/94#issuecomment-111630115.

puba commented 9 years ago

ay-ay 1:50 min delay on dgt xl clock

ifekali commented 9 years ago

Yes, I can reproduce. With logging on "info" the initial stall is still there.

BTW, we have a new bug introduced with latest version bump, the voices not announce two moves after user makes a move, but are silent when pico moves. (See the forums with user reporting it.)

The voices behave normally on stable branch (0.39).

On 13. jun. 2015, at 04.33, Shivkumar Shivaji notifications@github.com wrote:

I noticed this issue. Will solve this and a few other stability issues this weekend. Can you reproduce this issue if you change logging to info level?

  • Jonas Žnidaršič jonas@jonas.si

sshivaji commented 9 years ago

Initial stall should now be fixed in master. Clock detection logic was broken.

On Sat, Jun 13, 2015 at 3:00 PM, Izmet Fekali notifications@github.com wrote:

Yes, I can reproduce. With logging on "info" the initial stall is still there.

BTW, we have a new bug introduced with latest version bump, the voices not announce two moves after user makes a move, but are silent when pico moves. (See the forums with user reporting it.)

The voices behave normally on stable branch (0.39).

On 13. jun. 2015, at 04.33, Shivkumar Shivaji notifications@github.com wrote:

I noticed this issue. Will solve this and a few other stability issues this weekend. Can you reproduce this issue if you change logging to info level?

Jonas Žnidaršič jonas@jonas.si

— Reply to this email directly or view it on GitHub https://github.com/jromang/picochess/issues/94#issuecomment-111755315.

ifekali commented 9 years ago

Nope, it still stalls.

On 14. jun. 2015, at 03.20, Shivkumar Shivaji notifications@github.com wrote:

Initial stall should now be fixed in master. Clock detection logic was broken.

  • Jonas Žnidaršič jonas@jonas.si

ifekali commented 9 years ago

Puba reported it fixed too, but I'm guessing he simply got lucky one time. (This happens about once in twenty tries, the stall not being there.) The initial delay (stall) is still there on 0.41 master branch.

However, there is NO stall on stable 0.39 branch, I just checked. But apparently Shiv modified the STABLE branch (with remote play) 12 hours ago. Before that, I'm quite sure the stall was there before latest changes ...

On 14. jun. 2015, at 03.20, Shivkumar Shivaji notifications@github.com wrote:

Initial stall should now be fixed in master. Clock detection logic was broken.

  • Jonas Žnidaršič jonas@jonas.si

ifekali commented 9 years ago

Shiv, you modified STABLE branch (with first Remote play code). Indeed there is no more stall now there (showing 0.39 version still).

On 14. jun. 2015, at 03.20, Shivkumar Shivaji notifications@github.com wrote:

Initial stall should now be fixed in master.

  • Jonas Žnidaršič jonas@jonas.si

LocutusOfPenguin commented 9 years ago

for me the stall still exists, and its also there without board (=keyboard mode)

ifekali commented 9 years ago

Dang, the stall is indeed back on the stable branch. (0.39) This is weird.

On 14. jun. 2015, at 13.46, Jürgen Précour notifications@github.com wrote:

for me the stall still exists, and its also there without board (=keyboard mode)

  • Jonas Žnidaršič jonas@jonas.si

sshivaji commented 9 years ago

The stall is due to a threading issue and queue. Will commit fix once I figure out right solution. Yes, the issue is not easy to reproduce due to race condition.

On Jun 14, 2015, at 5:04 AM, Izmet Fekali notifications@github.com wrote:

Dang, the stall is indeed back on the stable branch. (0.39) This is weird.

On 14. jun. 2015, at 13.46, Jürgen Précour notifications@github.com wrote:

for me the stall still exists, and its also there without board (=keyboard mode)

  • Jonas Žnidaršič jonas@jonas.si

— Reply to this email directly or view it on GitHub.

sshivaji commented 9 years ago

Ok, fixed this one.

It is NOT a race condition like I assumed. Instead it turns out the geoip library we used takes a long while to respond and is not reliable. I moved that code to a thread and used a new library. Master branch should be good now.

On Sun, Jun 14, 2015 at 6:30 AM, Shivkumar Shivaji sshivaji@gmail.com wrote:

The stall is due to a threading issue and queue. Will commit fix once I figure out right solution. Yes, the issue is not easy to reproduce due to race condition.

On Jun 14, 2015, at 5:04 AM, Izmet Fekali notifications@github.com wrote:

Dang, the stall is indeed back on the stable branch. (0.39) This is weird.

On 14. jun. 2015, at 13.46, Jürgen Précour notifications@github.com wrote:

for me the stall still exists, and its also there without board (=keyboard mode)

Jonas Žnidaršič jonas@jonas.si

— Reply to this email directly or view it on GitHub https://github.com/jromang/picochess/issues/94#issuecomment-111817128.

puba commented 9 years ago

<a href="http://www.freesmileys.org/smileys.php" title="Smiley">Smiley