SAP CRM Fiori應用冗餘round trip的緣由分析

Sent: Wednesday, July 22, 2015 7:54 PM

有同事抱怨每次他們保存一個appointment時，除了正常的batch 操做外，還有3個莫名的read 操做。

The callstack clearly shows that the three roundtrips are NOT issued by customer extension, or else the customer js file could be observed in the callstack.

Set a breakpoint on the top most callstack, h function. Check the content of e.target.data:

This is actually the batch request payload which could be observed in Chrome network tab:

This finding gives me more confidence that these roundtrips are issued by framework, not standard or customer application code.

So I just continue debugging until I reach this suspicious stack:

in line 1957, this.bRefreshAfterChange = true.

However, in our internal system ( where everything works fine, there is no duplicate read operations ), this.bRefreshAfterChange = false, which has suppressed the refresh operation. This is the reason why the read operation could not be found in my internal system, since they are not executed at all. But in customer system, _isRefreshNeeded returns true, which leads to the execution of all subsequent read operations.

So why is this difference between two systems? In Chrome development tool, search the boolean variable name and we found one function setRefreshAfterChange defined for ODataModel. Just set a breakpoint in this method and re-launch the application in my internal system from beginning:

Breakpoint is triggered:

However, this line in customer system is missing, which is the root cause - our latest standard code didn't reach customer system.

出問題的系統上的標準代碼裏少了這一行，我在local的Eclipse裏試過，若是註釋掉，behavior就和出問題的系統上同樣，可以重現那三個多餘的讀操做了。

要獲取更多Jerry的原創文章，請關注公衆號"汪子熙":