F
Florian Fredegar Haftmann
Hi!
For a regression test setup on web interfaces, I use a test tool which
in a very heavywight fashion uses the COM interface to the Internet
Explorer (of course, from Python, ver 2.3.4, win32com build 163,
Windows XP, IExp ver 6.0.2800). During a session, many COM controls
are installed in order to trace actions inside the Internet Explorer.
After a certain amount (100?) of sessions (each about 2 until 5
minutes), allocating a new InternetExplorer control is not possible
any more, throwing an exception. The code snippet
[...]
try:
iexp = InternetExplorer()
except pythoncom_error:
...
[...]
(where InternetExplorer is the COM class materialized by makepy) then
leads to the except branch.
This "denial" state persists until a certain amount of time (may be
half an hour) has passed.
Its worth noting that each session is carried out sequentially in a
distinct process. I also made certain assurance tests on free memory
and CPU resources - there is definitvly no bottleneck. The issue
therefore seems to be a COM scale problem.
Has anyone experience with a comparable scale problem and knows how to
cirumvent it?
Any suggestion welcomed.
Thanks & Greetings
Florian Haftmann
For a regression test setup on web interfaces, I use a test tool which
in a very heavywight fashion uses the COM interface to the Internet
Explorer (of course, from Python, ver 2.3.4, win32com build 163,
Windows XP, IExp ver 6.0.2800). During a session, many COM controls
are installed in order to trace actions inside the Internet Explorer.
After a certain amount (100?) of sessions (each about 2 until 5
minutes), allocating a new InternetExplorer control is not possible
any more, throwing an exception. The code snippet
[...]
try:
iexp = InternetExplorer()
except pythoncom_error:
...
[...]
(where InternetExplorer is the COM class materialized by makepy) then
leads to the except branch.
This "denial" state persists until a certain amount of time (may be
half an hour) has passed.
Its worth noting that each session is carried out sequentially in a
distinct process. I also made certain assurance tests on free memory
and CPU resources - there is definitvly no bottleneck. The issue
therefore seems to be a COM scale problem.
Has anyone experience with a comparable scale problem and knows how to
cirumvent it?
Any suggestion welcomed.
Thanks & Greetings
Florian Haftmann